Skip to content

Latest commit

 

History

History
220 lines (182 loc) · 5.62 KB

NAF_version_3.1.md

File metadata and controls

220 lines (182 loc) · 5.62 KB
Raw

NAF version 3.1

This document describe the changes in NAF version 3.1 compared to NAF version 3. Each header provides information about a change.

  • date: June 26nd, 2020

Team

the following researchers contributed to this NAF version:

subtoken

We introduce a subtoken element which is a child of a wf element. The goal is to represent a subtoken of a word form element, typically used to represent compounds. The syntax of the id is TOKEN_ID.SUBTOKEN_ID, e.g., "w1.sub1", "w2.sub2", etc.

<!ELEMENT subtoken (#PCDATA)>
<!ATTLIST subtoken
          id ID #REQUIRED
          offset CDATA #REQUIRED
          length CDATA #REQUIRED>

wf

A subtoken is a child of a wf element, which is why the definition of the wf element has changed:

NAF version 3.0

<!ELEMENT wf (#PCDATA)>

NAF version 3.1

<!ELEMENT wf (#PCDATA|subtoken)*>

entity

In NAF version 3, an entities/entity element had a child called references. The references element has been removed. Instead, span is now directly a child of entities/entity

The type attribute of an entities/entity element is now optional.

predicate

The uri is removed from a predicate element.

role

Similarly, the role element values are defined in the externalReferences element in version 3.1 and no longer in the semRole attribute of the role element.

3.0

<!ELEMENT role (externalReferences|span)+>
<!ATTLIST role
          id ID #REQUIRED
          uri CDATA #IMPLIED
          confidence CDATA #IMPLIED
          semRole CDATA #REQUIRED>

3.1

<!ELEMENT role (externalReferences|span)+>
<!ATTLIST role
          id ID #REQUIRED
          confidence CDATA #IMPLIED
		  status CDATA #IMPLIED>

span

The following attribute is added to the span element:

  • status

The possible values of status are:

  • manual: manually added
  • system: a language system added the annotation
  • deprecated: the element is deprecated

We added this attribute to the following elements:

  • entities/entity
  • srl/predicate
  • srl/predicate/role
  • coref
  • coref/span
  • multiwords/mw

Version 3

<!-- SPAN ELEMENT -->
<!ELEMENT span (target)+>
<!ATTLIST span
          primary CDATA #IMPLIED>

Version 3.1

<!-- SPAN ELEMENT -->
<!ELEMENT span (target)+>
<!ATTLIST span
          primary CDATA #IMPLIED
		  status CDATA #IMPLIED>

component

A component element has not been changed, but we highlight the following naming convention:

If the component is part of a multiword, i.e., a multiwords/mw element, the syntax for the identifier is MW_ID.C_ID e.g., mw1.c1 and mw1.c2 for the two components of the multiwords/mw element with the identifier mw1. If the component is part of a compound, i.e., a terms/term element, the syntax for the identifier is T_ID.C_ID e.g., t1.c1 and t1.c2 for the two components of the terms/term element with the identifier t1.

term

The following attributes have been added to the term element:

  • component_of
  • compound_type

If the term is part of a multiword, i.e., there is a multiword/mw element that makes reference to it, then the component_of attribute can be used to make reference to the specific multiword.

The possible values of compound_type are:

  • endocentric: endocentric compound
  • exocentric: exocentric compoud

Version 3

<!ATTLIST term
          id ID #REQUIRED
          type CDATA #IMPLIED
          lemma CDATA #IMPLIED
          pos CDATA #IMPLIED
          morphofeat CDATA #IMPLIED
          netype CDATA #IMPLIED
          case CDATA #IMPLIED
          head CDATA #IMPLIED

Version 3.1

<!ATTLIST term
          id ID #REQUIRED
          type CDATA #IMPLIED
          lemma CDATA #IMPLIED
          pos CDATA #IMPLIED
          morphofeat CDATA #IMPLIED
          netype CDATA #IMPLIED
          case CDATA #IMPLIED
          head CDATA #IMPLIED
          component_of IDREF #IMPLIED
          compound_type CDATA #IMPLIED>

multiwords

Multiword expression are represented in the multiwords layer. Multi-word terms only have a component child in version 3.0. This component element can have multiple target elements, each referring to a component of the multi-word expression.

<!-- MULTIWORDS ELEMENT -->
<!ELEMENT multiwords (mw)+>

<!-- MW ELEMENT -->
<!--
    attributes of mw elements

    id: unique identifier (REQUIRED AND UNIQUE)

    lemma: lemma of the term (IMPLIED).

    pos: part of speech. (IMPLIED)

    morphofeat: morphosyntactic feature encoded as a single attribute. (IMPLIED)

    case: declension case (IMPLIED)

    type: phrasal, idiom
  -->

<!ELEMENT mw (component|externalReferences)+>
<!ATTLIST mw
          id ID #REQUIRED
          lemma CDATA #IMPLIED
          pos CDATA #IMPLIED
		  morphofeat CDATA #IMPLIED
		  case CDATA #IMPLIED
		  type CDATA #REQUIRED>

externalRef

We added the following attributes to the externalReferences/externalRef element:

  • timestamp: timestamp of adding externalRef

Version 3

<!ELEMENT externalRef (sentiment|externalRef)*>
<!ATTLIST externalRef
          resource CDATA #IMPLIED
          reference CDATA #REQUIRED
          reftype CDATA #IMPLIED
          status CDATA #IMPLIED
          source CDATA #IMPLIED
          confidence CDATA #IMPLIED>```

Version 3.1

<!ELEMENT externalRef (sentiment|externalRef)*>
<!ATTLIST externalRef
          resource CDATA #IMPLIED
          reference CDATA #REQUIRED
          reftype CDATA #IMPLIED
          status CDATA #IMPLIED
          source CDATA #IMPLIED
          confidence CDATA #IMPLIED
		  timestamp CDATA #IMPLIED>