Structural Approach to Enhancing WordNet with Conceptual Frame Semantics

Proceedings of Recent Advances in Natural Language Processing, pages 629–637, Varna, Bulgaria, Sep 2–4, 2019. https://doi.org/10.26615/978-954-452-056-4_074 629 A Structural Approach to Enhancing WordNet with Conceptual Frame Semantics Ivelina Stoyanova Institute for Bulgarian Language Bulgarian Academy of Sciences iva@dcl.bas.bg Svetlozara Leseva Institute for Bulgarian Language Bulgarian Academy of Sciences zarka@dcl.bas.bg Abstract This paper outlines procedures for enhanc- ing WordNet with conceptual information from FrameNet. The mapping of the two resources is non-trivial. We define a num- ber of techniques for the validation of the consistency of the mapping and the exten- sion of its coverage which make use of the structure of both resources and the system- atic relations between synsets in WordNet and between frames in FrameNet, as well as between synsets and frames). We present a case study on causativity, a relation which provides enhancement complementary to the one using hierarchi- cal relations, by means of linking in a sys- tematic way large parts of the lexicon. We show how consistency checks and denser relations may be implemented on the basis of this relation. We, then, propose new frames based on causative–inchoative correspondences and in conclusion touch on the possibilities for defining new frames based on the types of specialisation that takes place from parent to child synset. 1 Introduction The research presented in this paper aims at en- hancing WordNet with information about the con- ceptual structure of verbs based on the mappings between WordNet (WN) and FrameNet (FN). This information includes description of the conceptual elements that receive expression as verbs’ argu- ments and adjuncts as well as the selectional re- strictions imposed on these elements. Our approach relies on the structural features of each of the resources as well as on various kinds of linguistic analysis. The proposed enhancement is directed to: (a) improving the quality of exist- ing mappings; (b) expanding the mappings’ cov- erage; (c) enhancing the description of frames with additional information obtained from WN; (d) proposing structural improvements on the re- sources based on systemic features (e.g., causativ- ity), including the definition of new conceptual frames; and (e) suggesting further procedures for verification and improvements of precision. The described methodology is semi-automatic (automatic assignment complemented by manual verification at several stages), but its contribution consists in the devising of a set of procedures to a structured and consistent enrichment of the two resources, which presents an important step to- wards its automatisation. The enhanced resources will be made available to the research community. As the description of verbs’ conceptual structure is largely language independent, the enriched de- scription is applicable cross-linguistically. 2 Prerequisites and Motivation 2.1 Linguistic Resources WN (Miller, 1995; Fellbaum, 1998) is a large lexical database that represents comprehensively the conceptual and lexical knowledge in the form of a network whose nodes denote cognitive syn- onyms (synsets) interconnected through a number of conceptual-semantic and lexical relations. The main relation that determines WN’s structure (as reflected in the hierarchical treelike organisation of nouns and verbs) is a relation of inheritance of …

Content truncated to 500 words