Wikidata:Lexicographical data/Thematic relations

It is useful to store information about the semantic components of a word/phrase, whether these are roles to be taken on by other words/phrases or whether these are meanings of other words/phrases which contribute to the implications of said word/phrase, in a structured fashion. This has a number of benefits: it allows expressions for equivalent terms across languages to be matched in terms of the semantics as much as possible, and it aids the construction, for words in one language, of approximate phrases in other languages should those other languages not have exact equivalents. For Wikidata lexemes this information may be attached with has thematic relation (P9971).

(While the term "thematic role/relation" originates in the description of roles played by arguments to verbs and verb phrases, extending this term's umbrella to cover the semantic components of nouns/adjectives (and not just, say, Korean adjectives which do conjugate) is unlikely to introduce conflicts, especially when the latter primarily connect meaning components and the former often introduce (express/implied) place structure.)

The information in this page is largely inspired by the references presented at the bottom of this page, although as with most implementations within Wikidata lexicographical data is sure to deviate from them in some areas.

General

• A sense inherits the thematic relations from the hyperonym (P6593) values it bears.
  • Example: a hyponym of Hufschmied (L620801)-S1 will inherit "topic—Huf (L620800)-S1" from Hufschmied (L620801)-S1, which inherits "instrument—Metall (L48425)-S1" from Schmied (L500725)-S1.
• A sense may fill in some of the thematic relations it inherits with its own.
  • Example: dastum (L618346)-S1 specifies an "agent" and a "topic", but kistina (L40606), in addition to inheriting these, also specifies that the topic position is implied to be filled with kistin (L40605)-S1 and does not need to be expressed.

Verbs/verb phrases

Obligatory thematic relations may be attached via this property to senses on verb/verb phrase lexemes:

• The syntactic role of that relation may be provided as a object has role (P3831) qualifier.
  • Example: the entity doing the action of arbeiten (L408963)-S1 is noted as being a subject.
• The grammatical features that must be present on that relation may be provided as requires grammatical feature (P5713) qualifiers. These do not necessarily have to correspond to grammatical features on any forms.
  • Example: the entity who is targeted by döda (L38628)-S1 is marked as requiring the oblique case, despite Swedish common nouns lacking such a case, because a pronominal direct object must take that case rather than the nominative which common nouns take in this position.
• Optional thematic relations, as with other optional statements, may be marked with the qualifier nature of statement (P5102) optional (Q59864995). (There ought to be a reason for those relations' presence on that particular sense, perhaps if some general handling mechanism for them would not be adequate there.)
• Other implied relationships may be specified as qualifiers (object sense (P5980) preferred, but item for this sense (P5137) only if no direct expression exists).
  • Example: The action described at klimatanpassa (L619033)-S1 has the implied stimulus global uppvärmning (L618541)-S1. (If there were no Swedish lexeme describing global warming (Q7942), on which a P5137 value might have been given, then linking the item via P5137 might be acceptable.)

Nouns/adjectives

Thematic relations attached to noun/adjective senses qualify the meaning conveyed by their hyperonyms. (This goes hand-in-hand with the inheritance aspect listed in the "General" section of this page.)

References

• Dan Moldovan, Adriana Badulescu, Marta Tatu, Daniel Antohe, and Roxana Girju. 2004. Models for the semantic classification of noun phrases. In Proceedings of the HLT-NAACL Workshop on Computational Lexical Semantics, pages 60–67, Boston, Massachusetts, May. Association for Computational Linguistics.
• Claire Bonial, William Corvey, Martha Palmer, Volha V. Petukhova, and Harry Bunt. 2011. A Hierarchical Unification of LIRICS and VerbNet Semantic Roles. In 2011 IEEE Fifth International Conference on Semantic Computing, pages 483–489. September.
• Lasha Abzianidze and Johan Bos. 2017. Towards Universal Semantic Tagging. arXiv:1709.10381 (cs), September.
• Nathan Schneider, Vivek Srikumar, Jena D. Hwang, and Martha Palmer. 2015. A Hierarchy with, of, and for Preposition Supersenses. In Proceedings of The 9th Linguistic Annotation Workshop, pages 112–123, Denver, Colorado, USA, June. Association for Computational Linguistics.