Thematic Domain Group 7: Lexical Semantics

1
Thematic Domain Group 7 Lexical Semantics

• ISO TC 37/SC 4 N470
• ISO-TC37 meeting in Marrakech 25 May 2008
• MONICA MONACHINI
• ILC-CNR

2
Outline

• Situation Objectives
• Links with other ISO work LMF, TDGs
• Approaches
• Sources of info
• DC metamodel ISO 12620 samples of xml
  instantiations
• TDG7 task force and related activities
• Next work

3
Situation in TDG7 and Objectives

• New-born Thematic Domain Group on Lexical
  Semantics (Hong Kong ISO resolutions).
• Convenor MM
• Chairs Gil Francopoulo Nicoletta Calzolari
• Objective
• focus on lexical semantics for NLP lexicons
• establish a coherent family of low level
  standards (linguistic constants) for the lexical
  semantic profile
• augment the DCR with specific data categories to
  be used as elementary descriptors in combination
  with structural elements of LMF ISO 246132008
  when defining LMF-compliant lexicons

4
Links with other ISO activities

• interoperability between lexicon and syntactic
  and semantic annotation
• work shoulder to shoulder with
• LMF
• TDG5 (SynAF)
• TDG3 (SemAF)
• TDG6 ontologies
• ISO 12620 meta-model for DCs and ISOCat

5
TDG7 LMF semantic extension
DCR Semantic Profile
semantic relations
domain info
TDG6
ontological nodes
TDG5
TDG3
domain info
semantic relations
semantic roles
6
TDG7 LMF multilingual ext.
semantic relations
ontological nodes
TDG6
DCR Semantic Profile
7
ISO Approach

• Bottom-up procedure from best practices (à la
  Eagles)
• Collecting candidate semantic DCs
• Grouping, mapping, structuring
• Drafting a first set of definitions
• In TDG7, initially, focus on a set of candidate
  harmonized descriptors for semantic relations

8
Sources

• LIRICS set of agreed-on candidate semantic DCs
  for NLP lexicons ( SIMPLE harmonized
  plurilingual lexicons for 12 lang.s)
• KYOTO harmonization of synset semantic relations
  ( inter-WN intra-WN)
• but also
• BootStrep extension for the biology domain (e.g.
  OBO relations has_grain has_component )
• NEDO definition of harmonized predicate argument
  structures and semantic roles

9
Whats inherited from SIMPLE
10
SIMPLE set of semantic relations
60semanticrelations
11
LIRICS XML format
Conversion to ISO12620
12
ISO 12620 Data Category metamodel
13
Conceptual domain
A DC can be associated with a specific subset of
valid values described in the Descriptive
component (as a list of data categories)
14
Taxonomies of DCs
A DC can be linked to a broader concept, a more
general data category described in the
Descriptive component
DataCategory
Instrumental
DataCategory
Activity
hasABoaderDataCategory
DataCategory
DirectTelic
hasABoaderDataCategory
hasABoaderDataCategory
DataCategory
Telic
15
LIRICS XML format
has broader
has one value
Conversion to ISO12620
16
XML representation
17
XML representation
18
WordNet semantic relations mapping
A list of 85 sem.rels as a result of a mapping
of the KYOTOWordNet grid
Intra-WN
Inter-WN
19
TDG7 Community

• TDG7 mailing list set up to circulate the set of
  candidate DCs and allow discussion
• How to enlarge TDG7 task force?
• Link with the LMF mailing list
• Invite people to join from international projects
  we are involved in

20
Next activities

• Write procedures to convert the LIRICS stuff to
  the ISO-12620 format
• Refine the WordNet set of relations by adding
  definitions, examples,
• Provide xml instantiations of the agreed-on
  semantic DCs to upload in the new coming ISOCat
  tool
• Cooperate with ISOCat people
• Continue with other semantic info semantic
  relations (FrameNet), features, domain, link with
  ontologies

21
SIMPLE Ontology in OWL
CauseChangeofState
CauseChangeofState
CauseChangeofState
roastbeef
roast
createdBy