SIST ISO 24610-2:2013
(Main)Language resource management -- Feature structures -- Part 2: Feature system declaration
Language resource management -- Feature structures -- Part 2: Feature system declaration
This part of ISO 24610 provides a format to represent, store or exchange feature structures in natural language applications, for both annotation and production of linguistic data. It is ultimately designed to provide a computer format to define a type hierarchy and to declare the constraints that bear on a set of feature specifications and operations on feature structures, thus offering means to check the conformance of each feature structure with regards to a reference specification. Feature structures are an essential part of many linguistic formalisms as well as an underlying mechanism for representing the information consumed or produced by and for language engineering applications. A feature system declaration (FSD) is an auxiliary file used in conjunction with a certain type of text that makes use of fs (that is, feature structure) elements. The FSD serves four purposes. - It provides an encoding by which types and their subtyping and inheritance relationships can be introduced and defined, thus laying the basis for constructing a feature system. - It provides a mechanism by which the encoder can list all of the feature names and feature values and give a prose description as to what each represents. - It provides a mechanism by which type constraints can be declared, against which typed feature structures are validated relative to a given theory stated in typed feature logic. These constraints may involve constraints on the range of a feature's value, constraints on which features are permitted within certain types of feature structures, or constraints that prevent the co-occurrence of certain feature-value pairs. The source of these constraints is normally the empirical domain being modelled. - It provides a mechanism by which the encoder can define the intended interpretation of underspecified feature structures. This involves defining default values (whether literal or computed) for missing features. The scheme described in this part of ISO 24610 may be used to document any feature system, but is primarily intended for use with the typed feature structure representation defined in ISO 24610-1. The feature structure representations of ISO 24610-1 specify data structures that are subject to the typing conventions and constraints specified using ISO 24610-2. The feature structure representations of ISO 24610-1 are also used within some of the elements defined in ISO 24610-2.
Gestion des ressources langagières -- Structures de traits -- Partie 2: Déclaration de système de structures de traits
Upravljanje z jezikovnimi viri - Strukture lastnosti - 2. del: Deklaracija sistema lastnosti
Ta del standarda ISO 24610 zagotavlja format za predstavitev, shranjevanje in izmenjavo struktur lastnosti v aplikacijah za naravni jezik za označevanje in oblikovanje jezikovnih podatkov. Njegov glavni namen je zagotovitev računalniškega formata za določanje hierarhije tipov in opis omejitev, ki veljajo za sklop specifikacij lastnosti in operacij na strukturah lastnosti, in tako nudi način za preverjanje skladnosti vsake strukture lastnosti glede na referenčno specifikacijo. Strukture lastnosti so ključni sestavni del številnih jezikovnih formalizmov in osnovni mehanizmi za predstavitev informacij, ki jih predelajo ali oblikujejo aplikacije za jezikovno inženirstvo, ali ki so oblikovane za te aplikacije. Deklaracija sistema lastnosti (FSD) je pomožna datoteka, ki se uporablja z določeno vrsto besedila, in uporablja elemente struktur lastnosti. FSD ima štiri namene. – Zagotavlja kodiranje, prek katerega se lahko uvedejo in definirajo tipi in njihovi podtipi ter odnosi dedovanja, s čimer se postavijo temelji za gradnjo sistema lastnosti. – Zagotavlja mehanizem, s katerim lahko kodirnik oblikuje seznam vseh imen lastnosti in vrednosti lastnosti ter besedilu doda opis o tem, kaj predstavlja. – Zagotavlja mehanizem, s katerim se lahko navedejo omejitve tipa in s čimer se potrdijo tipizirane strukture lastnosti glede na dano teorijo, navedeno v logiki tipiziranih lastnosti. Te omejitve lahko vključujejo omejitve glede razpona vrednosti lastnosti, omejitve glede tega, katere lastnosti so dovoljene v določenih vrstah struktur lastnosti, ali omejitve, ki onemogočajo sopojavitev določenih parov lastnost-vrednost. Vir teh omejitev je običajno empirično modeliranje. – Zagotavlja mehanizem, s katerim lahko kodirnik določi želeno interpretacijo premalo specificiranih struktur lastnosti. To vključuje določanje privzetih vrednosti (dobesednih ali izračunanih) za manjkajoče vrednosti. Shema, ki jo opisuje ta del standarda ISO 24610, se lahko uporablja za dokumentiranje katerega koli sistema lastnosti, a je v prvi vrsti namenjena uporabi pri predstavitvi tipiziranih struktur lastnosti iz standarda ISO 24610-1. Predstavitve struktur lastnosti iz standarda ISO 24610-1 določajo strukture podatkov, za katere veljajo tipološke norme in omejitve, določene s pomočjo standarda ISO 24610-2. Predstavitve struktur lastnosti iz standarda ISO 24610-1 se uporabljajo tudi pri nekaterih elementih, določenih v standardu ISO 24610-2.
General Information
Buy Standard
Standards Content (Sample)
МЕЖДУНАРОДНЫЙ ISO
СТАНДАРТ 24610-2
Первое издание
2011-10-01
Управление языковыми ресурсами.
Структуры элементов.
Часть 2.
Декларация системы элементов
Language resource management. – Feature structures –
Part 2:
Feature system declaration
Ответственность за подготовку русской версии несѐт GOST R
(Российская Федерация) в соответствии со статьѐй 18.1 Устава ISO
Ссылочный номер
ISO 24610-2:2011(R)
©
ISO 2011
---------------------- Page: 1 ----------------------
ISO 24610-2:2011(R)
ДОКУМЕНТ ЗАЩИЩЁН АВТОРСКИМ ПРАВОМ
© ISO 2011
Все права сохраняются. Если не указано иное, никакую часть настоящей публикации нельзя копировать или использовать в
какой-либо форме или каким-либо электронным или механическим способом, включая фотокопии и микрофильмы, без
предварительного письменного согласия издателя.
ISO copyright office
Case postale 56 CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Опубликовано в Швейцарии
©
ii ISO 2011 – Все права сохраняются
---------------------- Page: 2 ----------------------
ISO 24610-2:2011(R)
Содержание Страница
Предисловие .iv
Введение .v
1 Область применения .1
2 Нормативные ссылки .1
3 Термины и определения .2
4 Общая структура стандарта .6
5 Базовые понятия .7
5.1 Рассматриваемые типизированные структуры элементов .7
5.2 Типы .8
5.3 Иерархии наследования типов . 11
5.4 Ограничения для типов . 12
5.5 Опциональные (стандартные) значения и недоопределение . 13
5.6 Категоризация . 14
6 Определение формальной правильности и адекватности . 16
6.1 Общее описание . 16
6.2 О стандарте ISO 24610 . 17
7 Система элементов для грамматики. 22
7.1 Общие сведения . 22
7.2 Выборочные FSD. 23
8 Декларация системы элементов . 27
8.1 Общие сведения . 27
8.2 Привязка текста к декларациям систем элементов . 28
8.3 Общая структура декларации системы элементов. 29
8.4 Декларации элементов . 31
8.5 Ограничения структуры элементов . 37
Приложение A (нормативное) Схема XML для структур элементов . 40
Приложение B (информативное) Детализированный пример . 50
Библиография . 54
©
ISO 2011 – Все права сохраняются iii
---------------------- Page: 3 ----------------------
ISO 24610-2:2011(R)
Предисловие
Международная организация по стандартизации (ISO) является всемирной федерацией национальных
организаций по стандартизации (комитетов-членов ISO). Разработка международных стандартов
обычно осуществляется техническими комитетами ISO. Каждый комитет-член, заинтересованный в
деятельности, для которой был создан технический комитет, имеет право быть представленным в этом
комитете. Международные правительственные и неправительственные организации, имеющие связь с
ISO, также принимают участие в работе. ISO работает в тесном сотрудничестве с Международной
электротехнической комиссией (IEC) по всем вопросам стандартизации в области электротехники.
Проекты международных стандартов разрабатываются согласно правилам, приведѐнным в Директивах
ISO/IEC, Часть 2.
Разработка международных стандартов является основной задачей технических комитетов. Проекты
международных стандартов, принятые техническими комитетами, рассылаются комитетам-членам на
голосование. Для публикации в качестве международного стандарта требуется одобрение не менее
75 % комитетов-членов, принявших участие в голосовании.
Принимается во внимание тот факт, что некоторые из элементов настоящей части стандарта ISO 9735
могут быть объектом патентных прав. ISO не принимает на себя обязательств по определению
отдельных или всех таких патентных прав.
ISO 24610-2 был подготовлен Техническим комитетом ISO/TC 37, Терминология и другие языковые и
информационные ресурсы, Подкомитет SC 4, Управление языковыми ресурсами.
В целом серия ISO 24610 состоит из следующих частей, объединѐнных общим названием Управление
языковыми ресурсами. Структуры элементов:
Часть 1. Представление структуры элементов
Часть 2. Декларация системы элементов
©
iv ISO 2011 – Все права сохраняются
---------------------- Page: 4 ----------------------
ISO 24610-2:2011(R)
Введение
ISO 24610 состоит из двух отдельных важных частей.
Часть 1, Представление структуры элементов, посвящена описанию структур, обеспечивающих
неформальное, но достаточно явное выражение их характеристик, а также описанию
представления структур элементов с использованием языка XML вообще и различных типов таких
структур, в частности. В этой части закладываются основы правильного форматирования
конструируемых XML-ссылок, обеспечивающих обмен структурами элементов (возможно, с
выделением типов) между приложениями.
Часть 2, Декларация системы элементов, предоставляет стандартный метод реализации
различных типов структур элементов в языковой среде XML: сначала путѐм определения
множества типов и их иерархии; затем посредством формулирования ограничений, касающихся
различных типов, на множестве элементов и их допустимых значений, и, наконец, путѐм введения
множества условий, касающихся надѐжности структур элементов в аспекте их использования в
конкретных приложениях, - особенно, в целях управления языковыми ресурсами.
Структура элементов – это структура данных общего назначения, которая идентифицирует и
группирует отдельные элементы посредством присваивания каждому из них конкретного значения.
Благодаря универсальности структур элементов они могут использоваться для представления самых
разных типов информации. Существующие связи между различными «порциями» информации и их
реализация в языке разметки образуют некоторый метаязык для представления контента
лингвистического характера. Более того, подобная реализация позволяет сформировать описание
множества элементов и значений, соответствующих конкретным типам и их ограничениям,
посредством декларирования системы элементов или с помощью других механизмов языка XML,
обсуждаемых в данной части ISO 24610.
Некоторые положения данной части заимствованы из ISO 24610-1:2006 в целях обеспечения полной
независимости части 2 от части 1.
©
ISO 2011 – Все права сохраняются v
---------------------- Page: 5 ----------------------
МЕЖДУНАРОДНЫЙ СТАНДАРТ ISO 24610-2:2011(R)
Управление языковыми ресурсами. Структуры элементов.
Часть 2.
Декларация системы элементов
1 Область применения
В данной части ISO 24610 предлагается формат представления, хранения и обмена для структур
элементов в прикладных системах, основанных на использовании естественного языка, как для
аннотирования, так и для формирования лингвистических данных. Основная цель состоит в том, чтобы
предложить такой формат машинной обработки, который позволяет определить иерархию типов и
декларировать ограничения, накладываемые на множество спецификаций элементов и на операции со
структурами элементов, обеспечивая таким образом средства контроля соответствия каждой
структуры элементов их базовой спецификации. Структуры элементов – это важнейшая часть многих
формализаций в лингвистике и основополагающий механизм представления информации,
используемой или порождаемой в приложениях, связанных с построением языковых систем.
Декларация системы элементов (FSD -feature system declaration) представляет собой вспомогательный
файл, относящийся к тексту конкретного типа, в рамках которого используются структурированные
элементы. Такая декларация служит четырѐм основным целям.
Обеспечивает кодирование, посредством которого могут вводиться и определяться типы и
подтипы, образующие основу для конструирования системы элементов;
Предоставляет механизм, с помощью которого кодировщик может сформировать список имѐн всех
элементов с соответствующими значениями и дать текстовое описание сущности каждого из них;
Реализует механизм декларирования разных типов ограничений, в соответствии с которыми
осуществляется контроль достоверности различных типов структур элементов на основе
использования теоретических принципов, установленных логикой выделения типов элементов;
этими ограничениями могут задаваться диапазон допустимых значений элемента, разрешѐнные
типы структур элементов или запрет на совместное вхождение в ту или иную структуру
определѐнных пар значений элементов; первоисточником таких ограничений обычно бывает
подлежащая моделированию предметная область;
Предоставляет механизм, посредством которого кодировщик может определять подразумеваемую
интерпретацию недоопределѐнных структур элементов: например, механизм определения
значений по умолчанию (литеральных или вычисляемых) для опущенных элементов.
Схема, описанная в данной части ISO 24610, может применяться для документирования любой
системы элементов, но предназначена, главным образом, для использования в рамках представлений
типизированных структур элементов, определѐнных в ISO 24610-1. Такие представления задают
структуры данных, подчиняющиеся условиям выделения типов и конкретным ограничениям,
определяемым с помощью ISO 24610-2. Представления структур элементов по ISO 24610-1
используются также применительно к некоторым элементам, определѐнным в ISO 24610-2.
2 Нормативные ссылки
Для применения данного документа необходимо обеспечение соответствия приведѐнным ниже
нормативным документам. Применительно к недатированным ссылочным документам (с плавающими
ссылками) действующим остаѐтся самое последнее издание нормативного документа.
©
ISO 2011 – Все права сохраняются 1
---------------------- Page: 6 ----------------------
ISO 24610-2:2011(R)
ISO 24610-1:2006, Управление языковыми ресурсами. Структуры элементов. Часть 1:
Представление структур элементов
ISO/IEC 19757-2, Информационные технологии. Язык определения схемы документа (DSDL).
Часть 2. Валидация на основе регулярной грамматики. RELAX NG
3 Термины и определения
Для целей данного документа используются термины и определения из стандарта ISO 19757-2, а
также терминология, приведѐнная ниже.
3.1
ограничение по допустимости
admissibility constraint
ограничение по разрешѐнным элементам
feature admissibility constraint
спецификация множества разрешѐнных элементов (3.2) и допустимых значений элементов (3.3),
ассоциируемая с конкретным типом (3.24)
3.2
разрешѐнный элемент
admissible feature
подходящий элемент
appropriate feature
элемент, для которого соответствующая структура элементов (3.14) определѐнного типа (3.24) может
нести в себе конкретное значение (3.17)
ПРИМЕЧАНИЕ В некоторых интерпретациях этот термин часто приобретает оттенок обязательности, то есть
считается, что структуры элементов конкретного типа должны содержать в себе значение для каждого
разрешѐнного элемента. Однако в нашем случае данный термин не предполагает обязательного присутствия
элемента.
3.3
разрешѐнное значение элемента
admissible feature value
допустимое значение
admissible value
ограничение по значениям
value restriction
ограничение по диапазону
range restriction
значение (3.17), которое должно быть отнесено к категории допустимых элементов (3.2) в
структурах элементов (3.14) данного типа (3.24)
3.4
атомарный тип
atomic type
пользовательский тип (3.24), который не имеет декларируемых или наследуемых допустимых
элементов (3.2)
3.5
множество с повторяющимися элементами
bag
мультимножество
multiset
триплет, образованный целым числом n, множеством S и функцией отображения целых чисел в
диапазоне от 1 до n, в элементы S
2
©
ISO 2011 – Все права сохраняются
---------------------- Page: 7 ----------------------
ISO 24610-2:2011(R)
ПРИМЕЧАНИЕ Множество с повторяющимися элементами – это промежуточный объект между обычным
множеством (как совокупностью неупорядоченных элементов) и списком (где отдельные элементы могут
встречаться многократно).
3.6
встроенный элемент
built-in
элемент, не определяемый пользователем, но могущий появиться вместо структуры элементов
(3.14), например, в качестве значения элемента (3.17)
ПРИМЕЧАНИЕ Встроенные элементы могут быть атомарными или составными. К первым относятся
численные, строковые, символьные и двоичные элементы; ко вторым - коллекции (3.7) и применяемые
логические операторы: например, дизъюнкция, отрицание и слияние (5.2.4).
3.7
коллекция
collection
значение элемента (3.17), содержащее совокупность возможных значений, которые представлены в
виде списка, обычного множества или множества с повторяющимися элементами (3.5)
3.8
ограничение
constraint
компонент спецификации, которая идентифицирует некоторую коллекцию структур элементов (3.14)
как неадекватную
ПРИМЕЧАНИЕ 1 Все ограничения по своей синтаксической форме импликативны, хотя некоторые из них
выделяются как ограничения по допустимости. См. адекватность (3.31) и 5.4. Все структуры элементов, которые
не исключены явным образом как неадекватные, считаются адекватными.
ПРИМЕЧАНИЕ 2 Структура элементов, не идентифицированная таким образом как не соответствующая
никакому из ограничений в системе элементов, считается адекватной.
3.9
значение по умолчанию, стандартное значение
default value
значение (3.17), присваиваемое элементу (3.12) в том случае, когда оно не определено
ПРИМЕР В датском языке при отсутствии явного указания грамматического рода ему присваивается значение
―мужской‖.
ПРИМЕЧАНИЕ Структура элементов не может содержать элементов, для которых не указано соответствующее
значение.
3.10
пустая структура элементов
empty feature structure
структура элементов (3.14), не содержащая никакой информации
ПРИМЕЧАНИЕ Пустая структура элементов категоризирует все другие структуры элементов.
3.11
расширение
extension
преобразование типа категоризации (3.21)
ПРИМЕЧАНИЕ Структура элементов F расширяет G тогда и только тогда, когда G категоризирует F.
©
ISO 2011 – Все права сохраняются 3
---------------------- Page: 8 ----------------------
ISO 24610-2:2011(R)
3.12
элемент
feature
свойство или аспект объекта, формально представляемые как функция, отображающая объект в его
соответствующее значение (3.17)
3.13
спецификация элементов
feature specification
связывание элемента (3.12) с его значением (3.17) в описании структуры элементов
3.14
структура элементов
feature structure
структура записей, которая ставит в соответствие каждой коллекции элементов одно значение (3.17)
ПРИМЕЧАНИЕ 1 Каждое значение представляет собой структуру элементов или более простой встроенный
элемент (3.6), такой как строка.
ПРИМЕЧАНИЕ 2 Структуры элементов частично упорядочены. Минимальными в этом упорядочении являются
пустые структуры элементов.
3.15
система элементов
feature system
иерархия типов (3.26), в которой каждый тип (3.24) ассоциируется с коллекцией ограничений по
допустимости (3.1) и импликативными ограничениями (3.18)
ПРИМЕЧАНИЕ ср. декларация типа (3.25)
3.16
декларация системы элементов
feature system declaration,
FSD
описание конкретной системы элементов (3.15)
3.17
значение для элемента
feature value
значение
value
объект или совокупность объектов, характеризующие некоторое свойство другого объекта
3.18
импликативное ограничение
implicational constraint
ограничение типа ―если G, то H,‖ где G и H – это структуры элементов (3.14)
ПРИМЕЧАНИЕ Такое ограничение идентифицирует любую структуру элементов F как неадекватную, когда
G категоризирует F, а F и H обычно не имеют адекватного расширения. См. категоризация (3.21) и 8.5. Часто
ограничение такого вида используется при обращении к импликативным ограничениям, которые одновременно не
являются ограничениями по допустимости.
3.19
интерпретация
interpretation
минимально информативное (то есть наиболее общее) расширение (3.11) структуры элементов
(3.14), которое совместимо с множеством ограничений, объявленным в декларации системы
элементов (3.16)
4
©
ISO 2011 – Все права сохраняются
---------------------- Page: 9 ----------------------
ISO 24610-2:2011(R)
3.20
частичный порядок
partial order
частично упорядоченное множество
partially ordered set
множество S, для которого определено отношение u на S S , которое 1) рефлексивно (для всех s S,
s u s), 2) антисимметрично (для всех p, q S, если p u q и q u p, то p q), и 3) транзитивно (для всех p, q,
r S, если p u q и q u r, то p u r)
ПРИМЕЧАНИЕ Множество целых чисел Z частично упорядочено, но дополнительно оно обладает свойством,
согласно которому, для каждого p, q Z, выполняется условие p u q или q u p. Этим свойством обладает не любой
частичный порядок. Например, такой частичный порядок, как таксономическая классификация организмов по
типам, родам и видам, указанным свойством не обладает; не обязательно обладают этим свойством также
иерархии типов. Типизированные структуры элементов системы не имеют этого свойства, если (a) данное
свойство присуще иерархии их типов, и (b) иерархия типов состоит из единственного типа либо каждый тип y
ограничен присутствием одного-единственного подходящего элемента.
3.21
категоризация
subsumption
свойство, связывающее две структуры элементов G и F таким образом, что G считается
принадлежащей F тогда и только тогда, когда F несѐт в себе всю информацию, которую содержит G
ПРИМЕЧАНИЕ Формальное определение представлено ниже, в 5.6.
3.22
подтип
subtype
тип (3.24), на который распространяются ограничения и соответствующие характеристики,
содержащиеся в другом типе
3.23
супертип, надтип
supertype
базовый тип
base type
тип (3.24), от которого другой тип наследует ограничения и соответствующие элементы
ПРИМЕЧАНИЕ s является подтипом t тогда и только тогда, когда t – супертип s. Каждый тип является
подтипом и супертипом самого себя.
3.24
семантический тип
semantic type
тип, характеризующий выражение, с помощью которого коллекция структур элементов (3.14)
различается как идентифицируемый и концептуально значимый класс
ПРИМЕЧАНИЕ Как это следует из имени семантический тип, типы, о которых идѐт речь в данной части
ISO 24610, не предназначены для различения структур элементов или их спецификаций по синтаксису.
3.25
декларация типа
type declaration
информационная структура, декларирующая супертипы (3.23), допустимые элементы (3.2),
значения допустимых элементов (3.3), ограничения по допустимости (3.1) и импликативные
ограничения (3.18) для данного типа (3.24)
ПРИМЕЧАНИЕ Ограничения, накладываемые на тип в результирующей системе элементов, это ограничения.
объявленные в декларации дополнительно к унаследованным от супертипов.
©
ISO 2011 – Все права сохраняются 5
---------------------- Page: 10 ----------------------
ISO 24610-2:2011(R)
3.26
иерархия типов
type hierarchy
частичный порядок (3.20) на множестве типов (3.24)
ПРИМЕЧАНИЯ См. ISO 24610-1:2006, Приложение C, Наследуемые иерархии типов.
3.27
типизированная структура элементов
typed feature structure, TFS
структура элементов (3.14), несущая в себе тип (3.24)
3.28
типизация
typing
присваивание семантического типа (3.24) встроенному элементу (3.6) либо структуре элементов
(3.14), атомарной или составной
ПРИМЕЧАНИЕ Семантические типы в системах элементов частично упорядочены и имеют множественные
отношения наследования.
3.29
недоопределение
underspecification
предоставление неполной информации о значении (3.17)
ПРИМЕЧАНИЕ Недоопределение обычно категоризирует одно значение из диапазона возможных значений,
которые могут быть сведены к единственному значению путѐм последовательного наложения ограничений. См.
категоризация (3.21).
3.30
формальная правильность
well-formedness
синтаксическое соответствие представления структуры элементов (3.14) и ISO 24610-1
3.31
адекватность
validity
соответствие типизированной структуры элементов (3.27) действующим ограничениям (3.8)
конкретной системы элементов (3.15)
ПРИМЕЧАНИЕ См. Раздел 6.
4 Общая структура стандарта
Основное содержание настоящего документа отражено в четырѐх разделах – 5, 6, 7 и 8.
В Разделе 5, Базовые понятия, рассматривается определение типизированных структур
элементов, и вводятся понятия атомарных и составных типов структур элементов, коллекций и
прочих операторов, могущих фигурировать в значениях элементов; затем описываются понятия
наследуемых типов, иерархий типов, ограничений типов, значений по умолчанию и
недоопределения, которые имеют важнейшее значение для конструирования систем элементов.
В Разделе 6, Определение формальной правильности и адекватности, обсуждаются условия
отмеченности и достоверности структур элементов.
6
©
ISO 2011 – Все права сохраняются
---------------------- Page: 11 ----------------------
ISO 24610-2:2011(R)
Раздел 7, Система элементов для грамматики, иллюстрирует способ определения типов с
использованием иерархии и ограничений типов, в рамках которых декларируются допустимые
элементы и значения для конкретных типов.
В Разделе 8, Декларация системы элементов, показывается, каким образом система элементов
может быть декларирована и преобразована в валидатор.
Эта главная часть документа включает в себя два приложения: Приложение A содержит the XML-
схему для данной части ISO 24610; Приложение B содержит развѐрнутый пример.
5 Базовые понятия
5.1 Рассматриваемые типизированные структуры элементов
Типизированные структуры элементов (TFS) вводятся как базовые записи для управления языковыми
ресурсами.
Для получения более подробной информации следует обратиться к ISO 24610-1:2006, 4.7,
Типизированные структуры элементов и Приложению C, Типизированные иерархии наследования.
В данном документе TFS определяется формально как кортеж на конечном множестве элементов Feat,
который состоит из коллекции X элементов, не входящих в структуру, и иерархии типов Type с
отношением u, где Type – это конечное множество типов, а отношение u определяет выделение
подтипов на множестве Type.
Структура элементов представляет собой кортеж , в котором:
a) Q – множество узлов,
b) γ ∈ Q – корневой узел структуры элементов,
c) θ : Q → Type является функцией частичного упорядочения, и
d) δ : Feat × Q → Q ∪ X – функция частичного означивания элементов, такая, что для всех q ∈ Q,
существует последовательность элементов F , ., F , в которой δ[F , . δ(F , γ) . ] q.
1 n n 1
Обозначение показывает узлы. Приведѐнное выше определение отличается от стандартного,
используемого в лингвистике и теории вычислительных систем, тем, что, во-первых, типизация
осуществляется частично, а не полностью (то есть типы определяются не для всех структур
элементов), и, во-вторых, значения элементов не обязательно должны представлять собой структуры
элементов; однако эти значения могут извлекаться из коллекции, отмеченной другими элементами
XML, - такими, как строковые, численные, символьные и двоичные (выше им соответствует
обозначение X). Следует заметить, что узлы типизируются, тогда как сами элементы - нет.
Приведѐнное ниже XML-представление структуры элементов считается формально правильным; в нѐм
атрибут ―тип‖ указывается для каждого из двух элементов .
ПРИМЕР Типизированная структура элементов:
had
©
ISO 2011 – Все права сохраняются 7
---------------------- Page: 12 ----------------------
ISO 24610-2:2011(R)
Имя элемента ORTH обозначает орфографию, то есть общепринятое написание слова или фразы.
Данное XML-представление показывает, каким образом определяются морфосинтаксические
характеристики английского слова ―had‖ как не вспомогательного глагола в прошедшем времени.
В альтернативной ―матричной‖ или ―AVM‖ нотации имена типов обычно пишутся строчными буквами,
иногда курсивом или текстовым типографским шрифтом; имена элементов пишутся заглавными
буквами, а строковые элементы заключаются в кавычки. Двоичные значения отмечаются знаками
―плюс‖ (+) или ―минус‖ (-). В данном документе эти соглашения тоже соблюдаются. Представленная
выше структура элементов должна при использовании матричной нотации выглядеть так, как показано
на Рисунке 1.
Рисунок 1 — Матричная нотация
5.2 Типы
5.2.1 Атомарные типы
Наряду со структурами с встроенными элементами (, , и ) могут
существовать структуры элементов, имеющие тип, но не имеющие элементов. Такие структуры
называются простыми, или атомарными структурами элементов, а типы, которые допускают
отсутствие элементов в декларации системы элементов (FSD), именуются атомарными типами.
В результате всегда имеется возможность декларирования новых атомарных типов и использования
их вместо вышеупомянутых встроенных элементов для задания простых значений. Например,
приведѐнная выше структура элементов при условии декларирования в FSD дополнительных типов
had, past и false могла бы быть представлена так, как показано ниже.
ПРИМЕР Альтернативная формулировка типизированной структуры элементов:
8
©
ISO 2011 – Все права сохраняются
---------------------- Page: 13 ----------------------
ISO 24610-2:2011(R)
Существует различие между двумя классами встроенных элементов: (строковый) и
(символьный), (двоичный), (численный). В качестве содержимого элемента
допустима любая строка, тогда как в элементах , и набор допустимых
значений строго ограничен. Для отражения такого различия значения членов последнего класса
определяются с использованием атрибута value. Тип , например, ассоциируется с четырьмя
значениями: true (истина), false (ложь), plus (эквивалент true) и minus (эквивалент false).
ПРИМЕЧАНИЕ В ISO 24610-1:2006 был введѐн тип binary (двоичный), но в схеме W3C XML (2001) он
называется Boolean (булев).
Задача кодировщика состоит в том, чтобы осуществить правильный выбор между кодированием
атомарных типов и встроенных элементов. В данной части ISO 24610 различие между двумя
вышеуказанными классами не проводится.
5.2.2 Составные типы
Типы, не являющиеся атомарными, называются составными. К ним относятся все типы,
декларируемые кодировщиком в FSD, где объявляются или наследуются допустимые элементы.
Элемент допустим для некоторого типа только в том случае, если структурам элементов данного типа
декларацией FSD разрешается принимать те или иные значения. Из этого не следует, что структуры
элементов не могут произвольно ассоциироваться с теми или иными типами независимо от их
элементного наполнения. Такое ассоциирование возможно, но проверяться на адекватность FSD
смогут лишь те структуры элементов, которые содержат только элементы, разрешѐнные какой-либо
FSD. Различие между адекватностью и формальной правильностью рассматривается более подробно
в Разделе 6.
Все типы, декларируемые пользователем (независимо от того, ат
...
SLOVENSKI STANDARD
SIST ISO 24610-2:2013
01-julij-2013
Upravljanje z jezikovnimi viri - Strukture lastnosti - 2. del: Deklaracija sistema
lastnosti
Language resource management -- Feature structures -- Part 2: Feature system
declaration
Gestion des ressources langagières -- Structures de traits -- Partie 2: Déclaration de
système de structures de traits
Ta slovenski standard je istoveten z: ISO 24610-2:2011
ICS:
01.020 Terminologija (načela in Terminology (principles and
koordinacija) coordination)
01.140.20 Informacijske vede Information sciences
35.240.30 Uporabniške rešitve IT v IT applications in information,
informatiki, dokumentiranju in documentation and
založništvu publishing
SIST ISO 24610-2:2013 en,fr,de
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
---------------------- Page: 1 ----------------------
SIST ISO 24610-2:2013
---------------------- Page: 2 ----------------------
SIST ISO 24610-2:2013
INTERNATIONAL ISO
STANDARD 24610-2
First edition
2011-10-01
Language resource management —
Feature structures —
Part 2:
Feature system declaration
Gestion des ressources langagières — Structures de traits —
Partie 2: Déclaration de système de structures de traits
Reference number
ISO 24610-2:2011(E)
©
ISO 2011
---------------------- Page: 3 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2011
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2011 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
Contents Page
Foreword . iv
Introduction . v
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 2
4 Overall structure . 5
5 Basic concepts . 6
5.1 Typed feature structures reviewed . 6
5.2 Types . 7
5.3 Type inheritance hierarchies . 9
5.4 Type constraints . 11
5.5 Optional (default) values and underspecification . 12
5.6 Subsumption . 12
6 Defining well-formedness versus validity. 14
6.1 Overview . 14
6.2 ISO 24610 . 14
7 A feature system for a grammar . 19
7.1 Overview . 19
7.2 Sample FSDs . 20
8 Declaration of a feature system . 23
8.1 Overview . 24
8.2 Linking a text to feature system declarations . 24
8.3 Overall structure of a feature system declaration . 25
8.4 Feature declarations . 27
8.5 Feature structure constraints . 33
Annex A (normative) XML schema for feature structures . 36
Annex B (informative) A complete example . 46
Bibliography . 50
© ISO 2011 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the
International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.
ISO 24610-2 was prepared by Technical Committee ISO/TC 37, Terminology and other language and content
resources, Subcommittee SC 4, Language resource management.
ISO 24610 consists of the following parts, under the general title Language resource management — Feature
structures:
Part 1: Feature structure representation
Part 2: Feature system declaration
iv © ISO 2011 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
Introduction
ISO 24610 is organized in two separate main parts.
Part 1, Feature structure representation, is dedicated to the description of feature structures, providing an
informal and yet explicit outline of their characteristics, as well as an XML-based structured way of
representing feature structures in general and typed feature structures in particular. It is designed to lay a
basis for constructing an XML-based reference format for exchanging (typed) feature structures between
applications.
Part 2, Feature system declaration, will provide an implementation standard for XML-based typed feature
structures, first by defining a set of types and their hierarchy, then by formulating type constraints on a set
of features and their respective admissible feature values and finally by introducing a set of validity
conditions on feature structures for particular applications, especially related to the goal of language
resource management.
A feature structure is a general-purpose data structure that identifies and groups together individual features
by assigning a particular value to each. Because of the generality of feature structures, they can be used to
represent many different kinds of information. Interrelations among various pieces of information and their
instantiation in markup provide a meta-language for representing linguistic content. Moreover, this
instantiation allows a specification of a set of features and values associated with specific types and their
restrictions, by means of feature system declarations, or other XML mechanisms to be discussed in this part
of ISO 24610.
Some of the statements here are copied from ISO 24610-1:2006 in order to make this part standalone without
referring to part 1.
© ISO 2011 – All rights reserved v
---------------------- Page: 7 ----------------------
SIST ISO 24610-2:2013
---------------------- Page: 8 ----------------------
SIST ISO 24610-2:2013
INTERNATIONAL STANDARD ISO 24610-2:2011(E)
Language resource management — Feature structures —
Part 2:
Feature system declaration
1 Scope
This part of ISO 24610 provides a format to represent, store or exchange feature structures in natural
language applications, for both annotation and production of linguistic data. It is ultimately designed to provide
a computer format to define a type hierarchy and to declare the constraints that bear on a set of feature
specifications and operations on feature structures, thus offering means to check the conformance of each
feature structure with regards to a reference specification. Feature structures are an essential part of many
linguistic formalisms as well as an underlying mechanism for representing the information consumed or
produced by and for language engineering applications.
A feature system declaration (FSD) is an auxiliary file used in conjunction with a certain type of text that
makes use of fs (that is, feature structure) elements. The FSD serves four purposes.
It provides an encoding by which types and their subtyping and inheritance relationships can be
introduced and defined, thus laying the basis for constructing a feature system.
It provides a mechanism by which the encoder can list all of the feature names and feature values and
give a prose description as to what each represents.
It provides a mechanism by which type constraints can be declared, against which typed feature
structures are validated relative to a given theory stated in typed feature logic. These constraints may
involve constraints on the range of a feature's value, constraints on which features are permitted within
certain types of feature structures, or constraints that prevent the co-occurrence of certain feature-value
pairs. The source of these constraints is normally the empirical domain being modelled.
It provides a mechanism by which the encoder can define the intended interpretation of underspecified
feature structures. This involves defining default values (whether literal or computed) for missing features.
The scheme described in this part of ISO 24610 may be used to document any feature system, but is primarily
intended for use with the typed feature structure representation defined in ISO 24610-1. The feature structure
representations of ISO 24610-1 specify data structures that are subject to the typing conventions and
constraints specified using ISO 24610-2. The feature structure representations of ISO 24610-1 are also used
within some of the elements defined in ISO 24610-2.
2 Normative references
The following referenced documents are indispensable for the application of this document. For dated
references, only the edition cited applies. For undated references, the latest edition of the referenced
document (including any amendments) applies.
ISO 24610-1:2006, Language resource management — Feature structures — Part 1: Feature structure
representation
ISO/IEC 19757-2, Information technology — Document Schema Definition Language (DSDL) — Part 2:
Regular-grammar-based validation — RELAX NG
© ISO 2011 – All rights reserved 1
---------------------- Page: 9 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO 19757-2 and the following apply.
3.1
admissibility constraint
feature admissibility constraint
specification of a set of admissible features (3.2) and admissible feature values (3.3) associated with a
specific type (3.24)
3.2
admissible feature
appropriate feature
feature which any feature structure (3.14) of a given type (3.24) may bear a value (3.17) for
NOTE This term is often interpreted elsewhere to mean obligatory, i.e. feature structures of the given type must bear
a value for every admissible feature. This term does not imply that the feature is obligatory here.
3.3
admissible feature value
admissible value
value restriction
range restriction
value (3.17) that the value of an admissible feature (3.2) must be subsumed by in feature structures (3.14)
of a given type (3.24)
3.4
atomic type
user-defined type (3.24) with no admissible features (3.2) declared or inherited
3.5
bag
multiset
triple of an integer n, a set S and a function that maps the integers in the range, 1 to n, to elements of S
NOTE A bag is halfway between a set (in that its elements are unordered) and a list (in that particular elements can
occur more than once).
3.6
built-in
non-user-defined element that may appear in place of a feature structure (3.14), for example, as a feature
value (3.17)
NOTE Built-ins can be atomic or complex. The atomic built-ins are numeric, string, symbol and binary. The complex
built-ins are collections (3.7) and applications of the operators, i.e. alternation, negation and merge (5.2.4).
3.7
collection
feature value (3.17) consisting of potentially many values, organized as a list, set or bag (3.5)
3.8
constraint
unit of specification that identifies some collection of feature structures (3.14) as invalid
NOTE 1 All constraints are implicational in their syntactic form, although some are distinguished as admissibility
constraints. See validity (3.31) and 5.4. All feature structures not explicitly excluded as invalid are considered to be valid.
NOTE 2 A feature structure that has not been so identified by any of the constraints in a feature system is considered
to be valid.
2 © ISO 2011 – All rights reserved
---------------------- Page: 10 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3.9
default value
value (3.17) otherwise assigned to a feature (3.12) when one is not specified
EXAMPLE Masculine is the default value of the grammatical gender in Dutch.
NOTE A feature structure may not bear a feature without a corresponding value.
3.10
empty feature structure
feature structure (3.14) that contains no information
NOTE An empty feature structure subsumes all other feature structures.
3.11
extension
converse of subsumption (3.21)
NOTE A feature structure F extends G if and only if G subsumes F.
3.12
feature
property or aspect of an entity that is formally represented as a function mapping the entity to a corresponding
value (3.17)
3.13
feature specification
pairing of a feature (3.12) with a value (3.17) in a feature structure description
3.14
feature structure
record structure that associates one value (3.17) to each of a collection of features
NOTE 1 Each value is either a feature structure or a simpler built-in (3.6) such as a string.
NOTE 2 Feature structures are partially ordered. The minimal feature structures in this ordering are the empty feature
structures.
3.15
feature system
type hierarchy (3.26) in which each type (3.24) has been associated with a collection of admissibility
constraints (3.1) and implicational constraints (3.18)
NOTE cf. type declaration (3.25)
3.16
feature system declaration
FSD
specification of a particular feature system (3.15)
3.17
feature value
value
entity or aggregation of entities that characterize some property or aspect of another entity
© ISO 2011 – All rights reserved 3
---------------------- Page: 11 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3.18
implicational constraint
constraint of the form, “if G, then H,” where G and H are feature structures (3.14)
NOTE This identifies any feature structure F as invalid for which G subsumes F, and yet F and H have no valid
extension in common. See subsumption (3.21) and 8.5. Often used to refer to implicational constraints that are not also
admissibility constraints.
3.19
interpretation
minimally informative (or equivalently, most general) extension (3.11) of a feature structure (3.14) that is
consistent with a set of constraints declared by an FSD (3.16)
3.20
partial order
partially ordered set
set S equipped with a relation over S S that is (1) reflexive (for all s S, s s), (2) anti-symmetric (for all p,
q S, if p q and q p, then p q), and (3) transitive (for all p, q, r S, if p q and q r, then p r)
NOTE The set of integers Z is partially ordered, but it has an additional property: for every p, q Z, either p q or
q p. Not all partial orders have this property. The taxonomical classification of organisms into phyla, genera and species,
for example, is a partial order that does not. Type hierarchies may not necessarily. The typed feature structures of a
feature system do not, unless (a) their type hierarchy does, and (b) either the type hierarchy has exactly one type, or every
y type is constrained to have exactly one appropriate feature.
3.21
subsumption
property that holds between two feature structures, G and F, such that G is said to subsume F if and only if F
carries all of the information with it that G does
NOTE A formal definition is provided in 5.6.
3.22
subtype
type (3.24) to which another type confers its constraints and appropriate features
3.23
supertype
base type
type (3.24) from which another type inherits constraints and appropriate features
NOTE s is a subtype of t iff t is a supertype of s. Every type is a subtype and supertype of itself.
3.24
semantic type
type
referring expression that distinguishes a collection of feature structures (3.14) as an identifiable and
conceptually significant class
NOTE As implied by the name semantic type, types in this part of ISO 24610 do not serve to distinguish feature
structures or their specifications syntactically.
3.25
type declaration
structure that declares the supertypes (3.23), admissible features (3.2), admissible feature values (3.3),
admissibility constraints (3.1) and implicational constraints (3.18) for a given type (3.24)
NOTE The constraints on a type in the resulting feature system are those that have been declared in its declaration,
in addition to those that it has inherited from its supertypes.
4 © ISO 2011 – All rights reserved
---------------------- Page: 12 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3.26
type hierarchy
partial order (3.20) over a set of types (3.24)
NOTE See ISO 24610-1:2006, Annex C, Type inheritance hierarchies.
3.27
typed feature structure
TFS
feature structure (3.14) that bears a type (3.24)
3.28
typing
assignment of a semantic type (3.24) to a built-in (3.6) or feature structure (3.14), either atomic or complex
NOTE Semantic types in feature systems are partially ordered, with multiple inheritance.
3.29
underspecification
provision of partial information about a value (3.17)
NOTE An underspecification generally subsumes one of a range of candidate values that could be resolved to a
single value through subsequent constraint resolution. See subsumption (3.21).
3.30
well-formedness
syntactic conformity of a feature structure (3.14) representation to ISO 24610-1
3.31
validity
conformity of a typed feature structure (3.27) to the constraints (3.8) of a particular feature system (3.15)
NOTE See Clause 6.
4 Overall structure
The main part of the document consists of four clauses: Clauses 5, 6, 7 and 8.
Clause 5, Basic concepts, reviews the definition of typed feature structures and the notions of atomic and
complex types, collections and other operators that may appear in feature values. It then describes the
notions of type inheritance hierarchies, type constraints, default values and underspecification that are
essential to the construction of feature systems.
Clause 6, Defining well-formedness versus validity, discusses the conditions of well-formedness and
validity.
Clause 7, A feature system for a grammar, illustrates how to define types with a type hierarchy and type
constraints which declare what features and values are admissible for specific types.
Finally, Clause 8, Declaration of a feature system, discusses how a feature system can be declared and
developed into a validator.
The main part of the document is followed by two annexes: Annex A contains the XML schema for this part of
ISO 24610; Annex B contains a complete example.
© ISO 2011 – All rights reserved 5
---------------------- Page: 13 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
5 Basic concepts
5.1 Typed feature structures reviewed
Typed feature structures (TFSs) are introduced as basic records for language resource management.
For more information, refer to ISO 24610-1:2006, 4.7, Typed feature structure, and Annex C, Type inheritance
hierarchies.
Here, a TFS is formally defined as a tuple over a finite set Feat of features, a collection X of
non-feature-structure elements, and a type hierarchy Type, , where Type is a finite set of types and is a
subtyping relation over Type.
A feature structure is a tuple , in which
a) Q is a set of nodes,
b) γ ∈ Q is the root node of the feature structure,
c) θ : Q → Type is a partial typing function, and
d) δ : Feat × Q → Q ∪ X is a partial feature value function,
such that, for all q ∈ Q, there exists a path of features F , ., F such that δ[F , . δ(F , γ) . ] q.
1 n n 1
elements denote nodes. This definition deviates from the standard one used in linguistics and theoretical
computer science in that (1) typing is partial, not total, i.e. not all feature structures have types, and (2) feature
values might not be feature structures, but instead be drawn from a collection denoted by other XML elements
such as string, numeric, symbol, and binary (the X above). Note that nodes are typed, but features themselves
are not.
The following XML representation of a feature structure is considered well-formed, where the attribute type is
assigned to each of the two elements.
EXAMPLE Typed feature structure:
had
The feature name ORTH above stands for orthography, the conventional written form of a word or phrase.
This XML representation shows how the morpho-syntactic features of an English word “had” are specified as
a past-tensed and non-auxiliary verb.
In the alternative, “matrix” or “AVM” notation, type names are conventionally in the lower-case, sometimes
italicized or in the text type font, feature names in the upper-case, and strings in quotes. Binary values are
6 © ISO 2011 – All rights reserved
---------------------- Page: 14 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
indicated with or . These conventions are followed in this document, too. The above feature structure would
be depicted in matrix notation as shown in Figure 1.
Figure 1 — Matrix notation
5.2 Types
5.2.1 Atomic types
Alongside the built-ins (, , and ), it is possible for a feature structure to
have a type but no features. These are called simple or atomic feature structures, and types that allow for no
features in their feature system declaration (FSD) are called atomic types.
There is, as a result, always the possibility of declaring new atomic types and using these instead of the
above-mentioned built-ins to specify simple values. The above feature structure, for example, could have
instead been rendered as follows, assuming the extra types had, past and false were declared in an FSD.
EXAMPLE Typed feature structure: alternative formulation
There is a difference also noticed between the two classes of built-ins: on the one hand, and
, and on the other. Any kind of string is permissible as the content of the
element, whereas a very restricted set of values is permissible in , and
elements. To reflect this difference, members of the latter class specify their values using the attribute value.
The type , for instance, is associated with four values: true, false, plus (equivalent to true) and minus
(equivalent to false).
NOTE ISO 24610-1:2006 introduced the type binary, but the W3C's XML schema (2001) names it boolean.
It is the duty of the encoder to choose between atomic-type encodings and built-in encodings consistently.
This part of ISO 24610 does not regard one as identical or even consistent with the other.
5.2.2 Complex types
Types that are not atomic are called complex. These include all of the types declared by the encoder in an
FSD that declare or inherit admissible features. A feature is only admissible to a type if feature structures of
that type are permitted by the FSD to have values for that feature. This does not mean that well-formed
feature structures cannot arbitrarily associate types with feature structures regardless of their featural content
– they can. But only those feature structures that use only admissible features to their type, as specified by
© ISO 2011 – All rights reserved 7
---------------------- Page: 15 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
some FSD, could be validated against that FSD. The distinction between validity and well-formedness is
further elaborated upon in Clause 6.
All user-declared types, no matter whether they are atomic or complex, are semantic, i.e. syntactically, they
look no different from each other, apart from the value of their type attribute. It is the role of a validator to
interpret the real significance of these types through enforcing restrictions on admissibility, restrictions on the
possible values that admissible features can have (), and other constraints that take the form of
logical implications. All of these are specified, for each type, in an FSD.
The built-ins defined by the ISO 24610-1:2006 feature structure representations (FSRs) standard are purely
syntactic. They can be used without declaration in an FSD, and they cannot be declared in an FSD. They can
appear in value range restrictions, or in implicational constraints, but they cannot have such restrictions (since
they have no admissible features) or constraints of their own.
5.2.3 Collections
Not all built-ins are as simple as those mentioned above, however. Some grammatical features such as
specifiers (SPR), complements (COMPS) and arguments (ARGS) are considered as having a list of
[10]
grammatical values, especially in Head-driven Phrase Structure Grammars (Pollard and Sag 1994 ; Sag,
[12]
Wasow, Bender 2003 ). For languages other than English, some of these features may take other kinds of
collections, namely sets or multisets, as their value. In a language (e.g. German, Korean or Japanese) that
allows a relatively free word order, the feature COMPS may be analysed as taking a set or multiset, instead of
a list, of complements. For more general applications, ISO 24610-1:2006 thus introduces sets and multisets
as well as lists as built-in ways of assembling complex feature values.
Collections (; ISO 24610-1:2006, 5.8, Collections as complex feature values) take the organization
(org) attribute, with the values “list”, “set” and “bag”. In lists, order and multi
...
INTERNATIONAL ISO
STANDARD 24610-2
First edition
2011-10-01
Language resource management —
Feature structures —
Part 2:
Feature system declaration
Gestion des ressources langagières — Structures de traits —
Partie 2: Déclaration de système de structures de traits
Reference number
ISO 24610-2:2011(E)
©
ISO 2011
---------------------- Page: 1 ----------------------
ISO 24610-2:2011(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2011
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2011 – All rights reserved
---------------------- Page: 2 ----------------------
ISO 24610-2:2011(E)
Contents Page
Foreword . iv
Introduction . v
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 2
4 Overall structure . 5
5 Basic concepts . 6
5.1 Typed feature structures reviewed . 6
5.2 Types . 7
5.3 Type inheritance hierarchies . 9
5.4 Type constraints . 11
5.5 Optional (default) values and underspecification . 12
5.6 Subsumption . 12
6 Defining well-formedness versus validity. 14
6.1 Overview . 14
6.2 ISO 24610 . 14
7 A feature system for a grammar . 19
7.1 Overview . 19
7.2 Sample FSDs . 20
8 Declaration of a feature system . 23
8.1 Overview . 24
8.2 Linking a text to feature system declarations . 24
8.3 Overall structure of a feature system declaration . 25
8.4 Feature declarations . 27
8.5 Feature structure constraints . 33
Annex A (normative) XML schema for feature structures . 36
Annex B (informative) A complete example . 46
Bibliography . 50
© ISO 2011 – All rights reserved iii
---------------------- Page: 3 ----------------------
ISO 24610-2:2011(E)
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the
International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.
ISO 24610-2 was prepared by Technical Committee ISO/TC 37, Terminology and other language and content
resources, Subcommittee SC 4, Language resource management.
ISO 24610 consists of the following parts, under the general title Language resource management — Feature
structures:
Part 1: Feature structure representation
Part 2: Feature system declaration
iv © ISO 2011 – All rights reserved
---------------------- Page: 4 ----------------------
ISO 24610-2:2011(E)
Introduction
ISO 24610 is organized in two separate main parts.
Part 1, Feature structure representation, is dedicated to the description of feature structures, providing an
informal and yet explicit outline of their characteristics, as well as an XML-based structured way of
representing feature structures in general and typed feature structures in particular. It is designed to lay a
basis for constructing an XML-based reference format for exchanging (typed) feature structures between
applications.
Part 2, Feature system declaration, will provide an implementation standard for XML-based typed feature
structures, first by defining a set of types and their hierarchy, then by formulating type constraints on a set
of features and their respective admissible feature values and finally by introducing a set of validity
conditions on feature structures for particular applications, especially related to the goal of language
resource management.
A feature structure is a general-purpose data structure that identifies and groups together individual features
by assigning a particular value to each. Because of the generality of feature structures, they can be used to
represent many different kinds of information. Interrelations among various pieces of information and their
instantiation in markup provide a meta-language for representing linguistic content. Moreover, this
instantiation allows a specification of a set of features and values associated with specific types and their
restrictions, by means of feature system declarations, or other XML mechanisms to be discussed in this part
of ISO 24610.
Some of the statements here are copied from ISO 24610-1:2006 in order to make this part standalone without
referring to part 1.
© ISO 2011 – All rights reserved v
---------------------- Page: 5 ----------------------
INTERNATIONAL STANDARD ISO 24610-2:2011(E)
Language resource management — Feature structures —
Part 2:
Feature system declaration
1 Scope
This part of ISO 24610 provides a format to represent, store or exchange feature structures in natural
language applications, for both annotation and production of linguistic data. It is ultimately designed to provide
a computer format to define a type hierarchy and to declare the constraints that bear on a set of feature
specifications and operations on feature structures, thus offering means to check the conformance of each
feature structure with regards to a reference specification. Feature structures are an essential part of many
linguistic formalisms as well as an underlying mechanism for representing the information consumed or
produced by and for language engineering applications.
A feature system declaration (FSD) is an auxiliary file used in conjunction with a certain type of text that
makes use of fs (that is, feature structure) elements. The FSD serves four purposes.
It provides an encoding by which types and their subtyping and inheritance relationships can be
introduced and defined, thus laying the basis for constructing a feature system.
It provides a mechanism by which the encoder can list all of the feature names and feature values and
give a prose description as to what each represents.
It provides a mechanism by which type constraints can be declared, against which typed feature
structures are validated relative to a given theory stated in typed feature logic. These constraints may
involve constraints on the range of a feature's value, constraints on which features are permitted within
certain types of feature structures, or constraints that prevent the co-occurrence of certain feature-value
pairs. The source of these constraints is normally the empirical domain being modelled.
It provides a mechanism by which the encoder can define the intended interpretation of underspecified
feature structures. This involves defining default values (whether literal or computed) for missing features.
The scheme described in this part of ISO 24610 may be used to document any feature system, but is primarily
intended for use with the typed feature structure representation defined in ISO 24610-1. The feature structure
representations of ISO 24610-1 specify data structures that are subject to the typing conventions and
constraints specified using ISO 24610-2. The feature structure representations of ISO 24610-1 are also used
within some of the elements defined in ISO 24610-2.
2 Normative references
The following referenced documents are indispensable for the application of this document. For dated
references, only the edition cited applies. For undated references, the latest edition of the referenced
document (including any amendments) applies.
ISO 24610-1:2006, Language resource management — Feature structures — Part 1: Feature structure
representation
ISO/IEC 19757-2, Information technology — Document Schema Definition Language (DSDL) — Part 2:
Regular-grammar-based validation — RELAX NG
© ISO 2011 – All rights reserved 1
---------------------- Page: 6 ----------------------
ISO 24610-2:2011(E)
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO 19757-2 and the following apply.
3.1
admissibility constraint
feature admissibility constraint
specification of a set of admissible features (3.2) and admissible feature values (3.3) associated with a
specific type (3.24)
3.2
admissible feature
appropriate feature
feature which any feature structure (3.14) of a given type (3.24) may bear a value (3.17) for
NOTE This term is often interpreted elsewhere to mean obligatory, i.e. feature structures of the given type must bear
a value for every admissible feature. This term does not imply that the feature is obligatory here.
3.3
admissible feature value
admissible value
value restriction
range restriction
value (3.17) that the value of an admissible feature (3.2) must be subsumed by in feature structures (3.14)
of a given type (3.24)
3.4
atomic type
user-defined type (3.24) with no admissible features (3.2) declared or inherited
3.5
bag
multiset
triple of an integer n, a set S and a function that maps the integers in the range, 1 to n, to elements of S
NOTE A bag is halfway between a set (in that its elements are unordered) and a list (in that particular elements can
occur more than once).
3.6
built-in
non-user-defined element that may appear in place of a feature structure (3.14), for example, as a feature
value (3.17)
NOTE Built-ins can be atomic or complex. The atomic built-ins are numeric, string, symbol and binary. The complex
built-ins are collections (3.7) and applications of the operators, i.e. alternation, negation and merge (5.2.4).
3.7
collection
feature value (3.17) consisting of potentially many values, organized as a list, set or bag (3.5)
3.8
constraint
unit of specification that identifies some collection of feature structures (3.14) as invalid
NOTE 1 All constraints are implicational in their syntactic form, although some are distinguished as admissibility
constraints. See validity (3.31) and 5.4. All feature structures not explicitly excluded as invalid are considered to be valid.
NOTE 2 A feature structure that has not been so identified by any of the constraints in a feature system is considered
to be valid.
2 © ISO 2011 – All rights reserved
---------------------- Page: 7 ----------------------
ISO 24610-2:2011(E)
3.9
default value
value (3.17) otherwise assigned to a feature (3.12) when one is not specified
EXAMPLE Masculine is the default value of the grammatical gender in Dutch.
NOTE A feature structure may not bear a feature without a corresponding value.
3.10
empty feature structure
feature structure (3.14) that contains no information
NOTE An empty feature structure subsumes all other feature structures.
3.11
extension
converse of subsumption (3.21)
NOTE A feature structure F extends G if and only if G subsumes F.
3.12
feature
property or aspect of an entity that is formally represented as a function mapping the entity to a corresponding
value (3.17)
3.13
feature specification
pairing of a feature (3.12) with a value (3.17) in a feature structure description
3.14
feature structure
record structure that associates one value (3.17) to each of a collection of features
NOTE 1 Each value is either a feature structure or a simpler built-in (3.6) such as a string.
NOTE 2 Feature structures are partially ordered. The minimal feature structures in this ordering are the empty feature
structures.
3.15
feature system
type hierarchy (3.26) in which each type (3.24) has been associated with a collection of admissibility
constraints (3.1) and implicational constraints (3.18)
NOTE cf. type declaration (3.25)
3.16
feature system declaration
FSD
specification of a particular feature system (3.15)
3.17
feature value
value
entity or aggregation of entities that characterize some property or aspect of another entity
© ISO 2011 – All rights reserved 3
---------------------- Page: 8 ----------------------
ISO 24610-2:2011(E)
3.18
implicational constraint
constraint of the form, “if G, then H,” where G and H are feature structures (3.14)
NOTE This identifies any feature structure F as invalid for which G subsumes F, and yet F and H have no valid
extension in common. See subsumption (3.21) and 8.5. Often used to refer to implicational constraints that are not also
admissibility constraints.
3.19
interpretation
minimally informative (or equivalently, most general) extension (3.11) of a feature structure (3.14) that is
consistent with a set of constraints declared by an FSD (3.16)
3.20
partial order
partially ordered set
set S equipped with a relation over S S that is (1) reflexive (for all s S, s s), (2) anti-symmetric (for all p,
q S, if p q and q p, then p q), and (3) transitive (for all p, q, r S, if p q and q r, then p r)
NOTE The set of integers Z is partially ordered, but it has an additional property: for every p, q Z, either p q or
q p. Not all partial orders have this property. The taxonomical classification of organisms into phyla, genera and species,
for example, is a partial order that does not. Type hierarchies may not necessarily. The typed feature structures of a
feature system do not, unless (a) their type hierarchy does, and (b) either the type hierarchy has exactly one type, or every
y type is constrained to have exactly one appropriate feature.
3.21
subsumption
property that holds between two feature structures, G and F, such that G is said to subsume F if and only if F
carries all of the information with it that G does
NOTE A formal definition is provided in 5.6.
3.22
subtype
type (3.24) to which another type confers its constraints and appropriate features
3.23
supertype
base type
type (3.24) from which another type inherits constraints and appropriate features
NOTE s is a subtype of t iff t is a supertype of s. Every type is a subtype and supertype of itself.
3.24
semantic type
type
referring expression that distinguishes a collection of feature structures (3.14) as an identifiable and
conceptually significant class
NOTE As implied by the name semantic type, types in this part of ISO 24610 do not serve to distinguish feature
structures or their specifications syntactically.
3.25
type declaration
structure that declares the supertypes (3.23), admissible features (3.2), admissible feature values (3.3),
admissibility constraints (3.1) and implicational constraints (3.18) for a given type (3.24)
NOTE The constraints on a type in the resulting feature system are those that have been declared in its declaration,
in addition to those that it has inherited from its supertypes.
4 © ISO 2011 – All rights reserved
---------------------- Page: 9 ----------------------
ISO 24610-2:2011(E)
3.26
type hierarchy
partial order (3.20) over a set of types (3.24)
NOTE See ISO 24610-1:2006, Annex C, Type inheritance hierarchies.
3.27
typed feature structure
TFS
feature structure (3.14) that bears a type (3.24)
3.28
typing
assignment of a semantic type (3.24) to a built-in (3.6) or feature structure (3.14), either atomic or complex
NOTE Semantic types in feature systems are partially ordered, with multiple inheritance.
3.29
underspecification
provision of partial information about a value (3.17)
NOTE An underspecification generally subsumes one of a range of candidate values that could be resolved to a
single value through subsequent constraint resolution. See subsumption (3.21).
3.30
well-formedness
syntactic conformity of a feature structure (3.14) representation to ISO 24610-1
3.31
validity
conformity of a typed feature structure (3.27) to the constraints (3.8) of a particular feature system (3.15)
NOTE See Clause 6.
4 Overall structure
The main part of the document consists of four clauses: Clauses 5, 6, 7 and 8.
Clause 5, Basic concepts, reviews the definition of typed feature structures and the notions of atomic and
complex types, collections and other operators that may appear in feature values. It then describes the
notions of type inheritance hierarchies, type constraints, default values and underspecification that are
essential to the construction of feature systems.
Clause 6, Defining well-formedness versus validity, discusses the conditions of well-formedness and
validity.
Clause 7, A feature system for a grammar, illustrates how to define types with a type hierarchy and type
constraints which declare what features and values are admissible for specific types.
Finally, Clause 8, Declaration of a feature system, discusses how a feature system can be declared and
developed into a validator.
The main part of the document is followed by two annexes: Annex A contains the XML schema for this part of
ISO 24610; Annex B contains a complete example.
© ISO 2011 – All rights reserved 5
---------------------- Page: 10 ----------------------
ISO 24610-2:2011(E)
5 Basic concepts
5.1 Typed feature structures reviewed
Typed feature structures (TFSs) are introduced as basic records for language resource management.
For more information, refer to ISO 24610-1:2006, 4.7, Typed feature structure, and Annex C, Type inheritance
hierarchies.
Here, a TFS is formally defined as a tuple over a finite set Feat of features, a collection X of
non-feature-structure elements, and a type hierarchy Type, , where Type is a finite set of types and is a
subtyping relation over Type.
A feature structure is a tuple , in which
a) Q is a set of nodes,
b) γ ∈ Q is the root node of the feature structure,
c) θ : Q → Type is a partial typing function, and
d) δ : Feat × Q → Q ∪ X is a partial feature value function,
such that, for all q ∈ Q, there exists a path of features F , ., F such that δ[F , . δ(F , γ) . ] q.
1 n n 1
elements denote nodes. This definition deviates from the standard one used in linguistics and theoretical
computer science in that (1) typing is partial, not total, i.e. not all feature structures have types, and (2) feature
values might not be feature structures, but instead be drawn from a collection denoted by other XML elements
such as string, numeric, symbol, and binary (the X above). Note that nodes are typed, but features themselves
are not.
The following XML representation of a feature structure is considered well-formed, where the attribute type is
assigned to each of the two elements.
EXAMPLE Typed feature structure:
had
The feature name ORTH above stands for orthography, the conventional written form of a word or phrase.
This XML representation shows how the morpho-syntactic features of an English word “had” are specified as
a past-tensed and non-auxiliary verb.
In the alternative, “matrix” or “AVM” notation, type names are conventionally in the lower-case, sometimes
italicized or in the text type font, feature names in the upper-case, and strings in quotes. Binary values are
6 © ISO 2011 – All rights reserved
---------------------- Page: 11 ----------------------
ISO 24610-2:2011(E)
indicated with or . These conventions are followed in this document, too. The above feature structure would
be depicted in matrix notation as shown in Figure 1.
Figure 1 — Matrix notation
5.2 Types
5.2.1 Atomic types
Alongside the built-ins (, , and ), it is possible for a feature structure to
have a type but no features. These are called simple or atomic feature structures, and types that allow for no
features in their feature system declaration (FSD) are called atomic types.
There is, as a result, always the possibility of declaring new atomic types and using these instead of the
above-mentioned built-ins to specify simple values. The above feature structure, for example, could have
instead been rendered as follows, assuming the extra types had, past and false were declared in an FSD.
EXAMPLE Typed feature structure: alternative formulation
There is a difference also noticed between the two classes of built-ins: on the one hand, and
, and on the other. Any kind of string is permissible as the content of the
element, whereas a very restricted set of values is permissible in , and
elements. To reflect this difference, members of the latter class specify their values using the attribute value.
The type , for instance, is associated with four values: true, false, plus (equivalent to true) and minus
(equivalent to false).
NOTE ISO 24610-1:2006 introduced the type binary, but the W3C's XML schema (2001) names it boolean.
It is the duty of the encoder to choose between atomic-type encodings and built-in encodings consistently.
This part of ISO 24610 does not regard one as identical or even consistent with the other.
5.2.2 Complex types
Types that are not atomic are called complex. These include all of the types declared by the encoder in an
FSD that declare or inherit admissible features. A feature is only admissible to a type if feature structures of
that type are permitted by the FSD to have values for that feature. This does not mean that well-formed
feature structures cannot arbitrarily associate types with feature structures regardless of their featural content
– they can. But only those feature structures that use only admissible features to their type, as specified by
© ISO 2011 – All rights reserved 7
---------------------- Page: 12 ----------------------
ISO 24610-2:2011(E)
some FSD, could be validated against that FSD. The distinction between validity and well-formedness is
further elaborated upon in Clause 6.
All user-declared types, no matter whether they are atomic or complex, are semantic, i.e. syntactically, they
look no different from each other, apart from the value of their type attribute. It is the role of a validator to
interpret the real significance of these types through enforcing restrictions on admissibility, restrictions on the
possible values that admissible features can have (), and other constraints that take the form of
logical implications. All of these are specified, for each type, in an FSD.
The built-ins defined by the ISO 24610-1:2006 feature structure representations (FSRs) standard are purely
syntactic. They can be used without declaration in an FSD, and they cannot be declared in an FSD. They can
appear in value range restrictions, or in implicational constraints, but they cannot have such restrictions (since
they have no admissible features) or constraints of their own.
5.2.3 Collections
Not all built-ins are as simple as those mentioned above, however. Some grammatical features such as
specifiers (SPR), complements (COMPS) and arguments (ARGS) are considered as having a list of
[10]
grammatical values, especially in Head-driven Phrase Structure Grammars (Pollard and Sag 1994 ; Sag,
[12]
Wasow, Bender 2003 ). For languages other than English, some of these features may take other kinds of
collections, namely sets or multisets, as their value. In a language (e.g. German, Korean or Japanese) that
allows a relatively free word order, the feature COMPS may be analysed as taking a set or multiset, instead of
a list, of complements. For more general applications, ISO 24610-1:2006 thus introduces sets and multisets
as well as lists as built-in ways of assembling complex feature values.
Collections (; ISO 24610-1:2006, 5.8, Collections as complex feature values) take the organization
(org) attribute, with the values “list”, “set” and “bag”. In lists, order and multiplicity of elements matter. In bags,
only multiplicity matters (these are often called multisets). In sets, neither order nor multiplicity matter.
For example, the feature ARGS of verbs can be represented by specifying the organization of as a list
of values, each of which is of type phrase.
EXAMPLE List value
put
8 © ISO 2011 – All rights reserved
---------------------- Page: 13 ----------------------
ISO 24610-2:2011(E)
Some would call the type of this collection list (phrase), but polymorphic lists are not yet supported in this part
of ISO 24610. This is equivalent to the following AVM notation, where NP stands for a feature structure of the
type phrase with a positive NOMINAL feature, namely a noun phrase, and PP, a feature structure of the type
phrase with a positive PREPOSITIONAL feature, namely a prepositional
...
SLOVENSKI STANDARD
SIST ISO 24610-2:2013
01-julij-2013
Upravljanje z jezikovnimi viri - Strukture lastnosti - 2. del: Deklaracija sistema
lastnosti
Language resource management -- Feature structures -- Part 2: Feature system
declaration
Gestion des ressources langagières -- Structures de traits -- Partie 2: Déclaration de
système de structures de traits
Ta slovenski standard je istoveten z: ISO 24610-2:2011
ICS:
01.140.20 Informacijske vede Information sciences
SIST ISO 24610-2:2013 en,fr,de
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.
---------------------- Page: 1 ----------------------
SIST ISO 24610-2:2013
---------------------- Page: 2 ----------------------
SIST ISO 24610-2:2013
INTERNATIONAL ISO
STANDARD 24610-2
First edition
2011-10-01
Language resource management —
Feature structures —
Part 2:
Feature system declaration
Gestion des ressources langagières — Structures de traits —
Partie 2: Déclaration de système de structures de traits
Reference number
ISO 24610-2:2011(E)
©
ISO 2011
---------------------- Page: 3 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
COPYRIGHT PROTECTED DOCUMENT
© ISO 2011
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form or by any means,
electronic or mechanical, including photocopying and microfilm, without permission in writing from either ISO at the address below or
ISO's member body in the country of the requester.
ISO copyright office
Case postale 56 CH-1211 Geneva 20
Tel. + 41 22 749 01 11
Fax + 41 22 749 09 47
E-mail copyright@iso.org
Web www.iso.org
Published in Switzerland
ii © ISO 2011 – All rights reserved
---------------------- Page: 4 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
Contents Page
Foreword . iv
Introduction . v
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 2
4 Overall structure . 5
5 Basic concepts . 6
5.1 Typed feature structures reviewed . 6
5.2 Types . 7
5.3 Type inheritance hierarchies . 9
5.4 Type constraints . 11
5.5 Optional (default) values and underspecification . 12
5.6 Subsumption . 12
6 Defining well-formedness versus validity. 14
6.1 Overview . 14
6.2 ISO 24610 . 14
7 A feature system for a grammar . 19
7.1 Overview . 19
7.2 Sample FSDs . 20
8 Declaration of a feature system . 23
8.1 Overview . 24
8.2 Linking a text to feature system declarations . 24
8.3 Overall structure of a feature system declaration . 25
8.4 Feature declarations . 27
8.5 Feature structure constraints . 33
Annex A (normative) XML schema for feature structures . 36
Annex B (informative) A complete example . 46
Bibliography . 50
© ISO 2011 – All rights reserved iii
---------------------- Page: 5 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
Foreword
ISO (the International Organization for Standardization) is a worldwide federation of national standards bodies
(ISO member bodies). The work of preparing International Standards is normally carried out through ISO
technical committees. Each member body interested in a subject for which a technical committee has been
established has the right to be represented on that committee. International organizations, governmental and
non-governmental, in liaison with ISO, also take part in the work. ISO collaborates closely with the
International Electrotechnical Commission (IEC) on all matters of electrotechnical standardization.
International Standards are drafted in accordance with the rules given in the ISO/IEC Directives, Part 2.
The main task of technical committees is to prepare International Standards. Draft International Standards
adopted by the technical committees are circulated to the member bodies for voting. Publication as an
International Standard requires approval by at least 75 % of the member bodies casting a vote.
Attention is drawn to the possibility that some of the elements of this document may be the subject of patent
rights. ISO shall not be held responsible for identifying any or all such patent rights.
ISO 24610-2 was prepared by Technical Committee ISO/TC 37, Terminology and other language and content
resources, Subcommittee SC 4, Language resource management.
ISO 24610 consists of the following parts, under the general title Language resource management — Feature
structures:
Part 1: Feature structure representation
Part 2: Feature system declaration
iv © ISO 2011 – All rights reserved
---------------------- Page: 6 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
Introduction
ISO 24610 is organized in two separate main parts.
Part 1, Feature structure representation, is dedicated to the description of feature structures, providing an
informal and yet explicit outline of their characteristics, as well as an XML-based structured way of
representing feature structures in general and typed feature structures in particular. It is designed to lay a
basis for constructing an XML-based reference format for exchanging (typed) feature structures between
applications.
Part 2, Feature system declaration, will provide an implementation standard for XML-based typed feature
structures, first by defining a set of types and their hierarchy, then by formulating type constraints on a set
of features and their respective admissible feature values and finally by introducing a set of validity
conditions on feature structures for particular applications, especially related to the goal of language
resource management.
A feature structure is a general-purpose data structure that identifies and groups together individual features
by assigning a particular value to each. Because of the generality of feature structures, they can be used to
represent many different kinds of information. Interrelations among various pieces of information and their
instantiation in markup provide a meta-language for representing linguistic content. Moreover, this
instantiation allows a specification of a set of features and values associated with specific types and their
restrictions, by means of feature system declarations, or other XML mechanisms to be discussed in this part
of ISO 24610.
Some of the statements here are copied from ISO 24610-1:2006 in order to make this part standalone without
referring to part 1.
© ISO 2011 – All rights reserved v
---------------------- Page: 7 ----------------------
SIST ISO 24610-2:2013
---------------------- Page: 8 ----------------------
SIST ISO 24610-2:2013
INTERNATIONAL STANDARD ISO 24610-2:2011(E)
Language resource management — Feature structures —
Part 2:
Feature system declaration
1 Scope
This part of ISO 24610 provides a format to represent, store or exchange feature structures in natural
language applications, for both annotation and production of linguistic data. It is ultimately designed to provide
a computer format to define a type hierarchy and to declare the constraints that bear on a set of feature
specifications and operations on feature structures, thus offering means to check the conformance of each
feature structure with regards to a reference specification. Feature structures are an essential part of many
linguistic formalisms as well as an underlying mechanism for representing the information consumed or
produced by and for language engineering applications.
A feature system declaration (FSD) is an auxiliary file used in conjunction with a certain type of text that
makes use of fs (that is, feature structure) elements. The FSD serves four purposes.
It provides an encoding by which types and their subtyping and inheritance relationships can be
introduced and defined, thus laying the basis for constructing a feature system.
It provides a mechanism by which the encoder can list all of the feature names and feature values and
give a prose description as to what each represents.
It provides a mechanism by which type constraints can be declared, against which typed feature
structures are validated relative to a given theory stated in typed feature logic. These constraints may
involve constraints on the range of a feature's value, constraints on which features are permitted within
certain types of feature structures, or constraints that prevent the co-occurrence of certain feature-value
pairs. The source of these constraints is normally the empirical domain being modelled.
It provides a mechanism by which the encoder can define the intended interpretation of underspecified
feature structures. This involves defining default values (whether literal or computed) for missing features.
The scheme described in this part of ISO 24610 may be used to document any feature system, but is primarily
intended for use with the typed feature structure representation defined in ISO 24610-1. The feature structure
representations of ISO 24610-1 specify data structures that are subject to the typing conventions and
constraints specified using ISO 24610-2. The feature structure representations of ISO 24610-1 are also used
within some of the elements defined in ISO 24610-2.
2 Normative references
The following referenced documents are indispensable for the application of this document. For dated
references, only the edition cited applies. For undated references, the latest edition of the referenced
document (including any amendments) applies.
ISO 24610-1:2006, Language resource management — Feature structures — Part 1: Feature structure
representation
ISO/IEC 19757-2, Information technology — Document Schema Definition Language (DSDL) — Part 2:
Regular-grammar-based validation — RELAX NG
© ISO 2011 – All rights reserved 1
---------------------- Page: 9 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO 19757-2 and the following apply.
3.1
admissibility constraint
feature admissibility constraint
specification of a set of admissible features (3.2) and admissible feature values (3.3) associated with a
specific type (3.24)
3.2
admissible feature
appropriate feature
feature which any feature structure (3.14) of a given type (3.24) may bear a value (3.17) for
NOTE This term is often interpreted elsewhere to mean obligatory, i.e. feature structures of the given type must bear
a value for every admissible feature. This term does not imply that the feature is obligatory here.
3.3
admissible feature value
admissible value
value restriction
range restriction
value (3.17) that the value of an admissible feature (3.2) must be subsumed by in feature structures (3.14)
of a given type (3.24)
3.4
atomic type
user-defined type (3.24) with no admissible features (3.2) declared or inherited
3.5
bag
multiset
triple of an integer n, a set S and a function that maps the integers in the range, 1 to n, to elements of S
NOTE A bag is halfway between a set (in that its elements are unordered) and a list (in that particular elements can
occur more than once).
3.6
built-in
non-user-defined element that may appear in place of a feature structure (3.14), for example, as a feature
value (3.17)
NOTE Built-ins can be atomic or complex. The atomic built-ins are numeric, string, symbol and binary. The complex
built-ins are collections (3.7) and applications of the operators, i.e. alternation, negation and merge (5.2.4).
3.7
collection
feature value (3.17) consisting of potentially many values, organized as a list, set or bag (3.5)
3.8
constraint
unit of specification that identifies some collection of feature structures (3.14) as invalid
NOTE 1 All constraints are implicational in their syntactic form, although some are distinguished as admissibility
constraints. See validity (3.31) and 5.4. All feature structures not explicitly excluded as invalid are considered to be valid.
NOTE 2 A feature structure that has not been so identified by any of the constraints in a feature system is considered
to be valid.
2 © ISO 2011 – All rights reserved
---------------------- Page: 10 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3.9
default value
value (3.17) otherwise assigned to a feature (3.12) when one is not specified
EXAMPLE Masculine is the default value of the grammatical gender in Dutch.
NOTE A feature structure may not bear a feature without a corresponding value.
3.10
empty feature structure
feature structure (3.14) that contains no information
NOTE An empty feature structure subsumes all other feature structures.
3.11
extension
converse of subsumption (3.21)
NOTE A feature structure F extends G if and only if G subsumes F.
3.12
feature
property or aspect of an entity that is formally represented as a function mapping the entity to a corresponding
value (3.17)
3.13
feature specification
pairing of a feature (3.12) with a value (3.17) in a feature structure description
3.14
feature structure
record structure that associates one value (3.17) to each of a collection of features
NOTE 1 Each value is either a feature structure or a simpler built-in (3.6) such as a string.
NOTE 2 Feature structures are partially ordered. The minimal feature structures in this ordering are the empty feature
structures.
3.15
feature system
type hierarchy (3.26) in which each type (3.24) has been associated with a collection of admissibility
constraints (3.1) and implicational constraints (3.18)
NOTE cf. type declaration (3.25)
3.16
feature system declaration
FSD
specification of a particular feature system (3.15)
3.17
feature value
value
entity or aggregation of entities that characterize some property or aspect of another entity
© ISO 2011 – All rights reserved 3
---------------------- Page: 11 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3.18
implicational constraint
constraint of the form, “if G, then H,” where G and H are feature structures (3.14)
NOTE This identifies any feature structure F as invalid for which G subsumes F, and yet F and H have no valid
extension in common. See subsumption (3.21) and 8.5. Often used to refer to implicational constraints that are not also
admissibility constraints.
3.19
interpretation
minimally informative (or equivalently, most general) extension (3.11) of a feature structure (3.14) that is
consistent with a set of constraints declared by an FSD (3.16)
3.20
partial order
partially ordered set
set S equipped with a relation over S S that is (1) reflexive (for all s S, s s), (2) anti-symmetric (for all p,
q S, if p q and q p, then p q), and (3) transitive (for all p, q, r S, if p q and q r, then p r)
NOTE The set of integers Z is partially ordered, but it has an additional property: for every p, q Z, either p q or
q p. Not all partial orders have this property. The taxonomical classification of organisms into phyla, genera and species,
for example, is a partial order that does not. Type hierarchies may not necessarily. The typed feature structures of a
feature system do not, unless (a) their type hierarchy does, and (b) either the type hierarchy has exactly one type, or every
y type is constrained to have exactly one appropriate feature.
3.21
subsumption
property that holds between two feature structures, G and F, such that G is said to subsume F if and only if F
carries all of the information with it that G does
NOTE A formal definition is provided in 5.6.
3.22
subtype
type (3.24) to which another type confers its constraints and appropriate features
3.23
supertype
base type
type (3.24) from which another type inherits constraints and appropriate features
NOTE s is a subtype of t iff t is a supertype of s. Every type is a subtype and supertype of itself.
3.24
semantic type
type
referring expression that distinguishes a collection of feature structures (3.14) as an identifiable and
conceptually significant class
NOTE As implied by the name semantic type, types in this part of ISO 24610 do not serve to distinguish feature
structures or their specifications syntactically.
3.25
type declaration
structure that declares the supertypes (3.23), admissible features (3.2), admissible feature values (3.3),
admissibility constraints (3.1) and implicational constraints (3.18) for a given type (3.24)
NOTE The constraints on a type in the resulting feature system are those that have been declared in its declaration,
in addition to those that it has inherited from its supertypes.
4 © ISO 2011 – All rights reserved
---------------------- Page: 12 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
3.26
type hierarchy
partial order (3.20) over a set of types (3.24)
NOTE See ISO 24610-1:2006, Annex C, Type inheritance hierarchies.
3.27
typed feature structure
TFS
feature structure (3.14) that bears a type (3.24)
3.28
typing
assignment of a semantic type (3.24) to a built-in (3.6) or feature structure (3.14), either atomic or complex
NOTE Semantic types in feature systems are partially ordered, with multiple inheritance.
3.29
underspecification
provision of partial information about a value (3.17)
NOTE An underspecification generally subsumes one of a range of candidate values that could be resolved to a
single value through subsequent constraint resolution. See subsumption (3.21).
3.30
well-formedness
syntactic conformity of a feature structure (3.14) representation to ISO 24610-1
3.31
validity
conformity of a typed feature structure (3.27) to the constraints (3.8) of a particular feature system (3.15)
NOTE See Clause 6.
4 Overall structure
The main part of the document consists of four clauses: Clauses 5, 6, 7 and 8.
Clause 5, Basic concepts, reviews the definition of typed feature structures and the notions of atomic and
complex types, collections and other operators that may appear in feature values. It then describes the
notions of type inheritance hierarchies, type constraints, default values and underspecification that are
essential to the construction of feature systems.
Clause 6, Defining well-formedness versus validity, discusses the conditions of well-formedness and
validity.
Clause 7, A feature system for a grammar, illustrates how to define types with a type hierarchy and type
constraints which declare what features and values are admissible for specific types.
Finally, Clause 8, Declaration of a feature system, discusses how a feature system can be declared and
developed into a validator.
The main part of the document is followed by two annexes: Annex A contains the XML schema for this part of
ISO 24610; Annex B contains a complete example.
© ISO 2011 – All rights reserved 5
---------------------- Page: 13 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
5 Basic concepts
5.1 Typed feature structures reviewed
Typed feature structures (TFSs) are introduced as basic records for language resource management.
For more information, refer to ISO 24610-1:2006, 4.7, Typed feature structure, and Annex C, Type inheritance
hierarchies.
Here, a TFS is formally defined as a tuple over a finite set Feat of features, a collection X of
non-feature-structure elements, and a type hierarchy Type, , where Type is a finite set of types and is a
subtyping relation over Type.
A feature structure is a tuple , in which
a) Q is a set of nodes,
b) γ ∈ Q is the root node of the feature structure,
c) θ : Q → Type is a partial typing function, and
d) δ : Feat × Q → Q ∪ X is a partial feature value function,
such that, for all q ∈ Q, there exists a path of features F , ., F such that δ[F , . δ(F , γ) . ] q.
1 n n 1
elements denote nodes. This definition deviates from the standard one used in linguistics and theoretical
computer science in that (1) typing is partial, not total, i.e. not all feature structures have types, and (2) feature
values might not be feature structures, but instead be drawn from a collection denoted by other XML elements
such as string, numeric, symbol, and binary (the X above). Note that nodes are typed, but features themselves
are not.
The following XML representation of a feature structure is considered well-formed, where the attribute type is
assigned to each of the two elements.
EXAMPLE Typed feature structure:
had
The feature name ORTH above stands for orthography, the conventional written form of a word or phrase.
This XML representation shows how the morpho-syntactic features of an English word “had” are specified as
a past-tensed and non-auxiliary verb.
In the alternative, “matrix” or “AVM” notation, type names are conventionally in the lower-case, sometimes
italicized or in the text type font, feature names in the upper-case, and strings in quotes. Binary values are
6 © ISO 2011 – All rights reserved
---------------------- Page: 14 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
indicated with or . These conventions are followed in this document, too. The above feature structure would
be depicted in matrix notation as shown in Figure 1.
Figure 1 — Matrix notation
5.2 Types
5.2.1 Atomic types
Alongside the built-ins (, , and ), it is possible for a feature structure to
have a type but no features. These are called simple or atomic feature structures, and types that allow for no
features in their feature system declaration (FSD) are called atomic types.
There is, as a result, always the possibility of declaring new atomic types and using these instead of the
above-mentioned built-ins to specify simple values. The above feature structure, for example, could have
instead been rendered as follows, assuming the extra types had, past and false were declared in an FSD.
EXAMPLE Typed feature structure: alternative formulation
There is a difference also noticed between the two classes of built-ins: on the one hand, and
, and on the other. Any kind of string is permissible as the content of the
element, whereas a very restricted set of values is permissible in , and
elements. To reflect this difference, members of the latter class specify their values using the attribute value.
The type , for instance, is associated with four values: true, false, plus (equivalent to true) and minus
(equivalent to false).
NOTE ISO 24610-1:2006 introduced the type binary, but the W3C's XML schema (2001) names it boolean.
It is the duty of the encoder to choose between atomic-type encodings and built-in encodings consistently.
This part of ISO 24610 does not regard one as identical or even consistent with the other.
5.2.2 Complex types
Types that are not atomic are called complex. These include all of the types declared by the encoder in an
FSD that declare or inherit admissible features. A feature is only admissible to a type if feature structures of
that type are permitted by the FSD to have values for that feature. This does not mean that well-formed
feature structures cannot arbitrarily associate types with feature structures regardless of their featural content
– they can. But only those feature structures that use only admissible features to their type, as specified by
© ISO 2011 – All rights reserved 7
---------------------- Page: 15 ----------------------
SIST ISO 24610-2:2013
ISO 24610-2:2011(E)
some FSD, could be validated against that FSD. The distinction between validity and well-formedness is
further elaborated upon in Clause 6.
All user-declared types, no matter whether they are atomic or complex, are semantic, i.e. syntactically, they
look no different from each other, apart from the value of their type attribute. It is the role of a validator to
interpret the real significance of these types through enforcing restrictions on admissibility, restrictions on the
possible values that admissible features can have (), and other constraints that take the form of
logical implications. All of these are specified, for each type, in an FSD.
The built-ins defined by the ISO 24610-1:2006 feature structure representations (FSRs) standard are purely
syntactic. They can be used without declaration in an FSD, and they cannot be declared in an FSD. They can
appear in value range restrictions, or in implicational constraints, but they cannot have such restrictions (since
they have no admissible features) or constraints of their own.
5.2.3 Collections
Not all built-ins are as simple as those mentioned above, however. Some grammatical features such as
specifiers (SPR), complements (COMPS) and arguments (ARGS) are considered as having a list of
[10]
grammatical values, especially in Head-driven Phrase Structure Grammars (Pollard and Sag 1994 ; Sag,
[12]
Wasow, Bender 2003 ). For languages other than English, some of these features may take other kinds of
collections, namely sets or multisets, as their value. In a language (e.g. German, Korean or Japanese) that
allows a relatively free word order, the feature COMPS may be analysed as taking a set or multiset, instead of
a list, of complements. For more general applications, ISO 24610-1:2006 thus introduces sets and multisets
as well as lists as built-in ways of assembling complex feature values.
Collections (; ISO 24610-1:2006, 5.8, Collections as complex feature values) take the organization
(org) attribute, with the values “list”, “set” and “bag”. In lists, order and multiplicity of elements matter. In bags,
only multiplicity matters (these are often called multisets). In sets, neither order nor multiplicity matter.
For example, the feature ARGS of verbs can be repres
...
Questions, Comments and Discussion
Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.