Modelling Frequency, Attestation, and Corpus-Based Information with OntoLex-FrAC

Proceedings of the 29th International Conference on Computational Linguistics, pages 4018–4027 October 12–17, 2022. 4018 Modelling Frequency, Attestation, and Corpus-Based Information with OntoLex-FrAC Christian Chiarcos1,2 and Elena-Simona Apostol3,4 and Besim Kabashi5 and Ciprian-Octavian Truic˘a3,4 1Applied Computational Linguistics, Goethe University Frankfurt, Germany 2Institute for Digital Humanities, University of Cologne, Germany 3Department of Information Technology, Uppsala University, Sweden 4Faculty of Automatic Control and Computers, University Politehnica of Bucharest, Romania 5Computational and Corpus Linguistics, University of Erlangen-Nuremberg, Germany chiarcos@cs.uni-frankfurt.de, elena-simona.apostol@it.uu.se besim.kabashi@fau.de, ciprian-octavian.truica@it.uu.se Abstract OntoLex-Lemon has become a de facto stan- dard for lexical resources in the web of data. This paper provides the first overall descrip- tion of the emerging OntoLex module for Fre- quency, Attestations, and Corpus-Based Infor- mation (OntoLex-FrAC) that is intended to complement OntoLex-Lemon with the neces- sary vocabulary to represent major types of information found in or automatically derived from corpora, for applications in both language technology and the language sciences. 1 Background The OntoLex-Lemon vocabulary has become the dominant vocabulary for modelling machine- readable dictionaries on the web of data, i.e., by means of RDF. And indeed, publishing lexical re- sources in RDF has a number of advantages, includ- ing the ease of integration of dictionary information not only with ontologies and knowledge graphs (this was the original domain of application), but also with other lexical data. Figure 1 illustrates the OntoLex-Lemon core vocabulary. Primary data structures are ontolex: LexicalEntry (lexeme), ontolex:Form (word form), ontolex:LexicalSense (word Figure 1: OntoLex-Lemon core module sense), and ontolex:LexicalConcept (lexicalization-independent concept), so that lexi- cal entries can be described, but also fine-grained differences in meaning and surface form. While these aspects are advanced, stable and widely used, there is no complete module described in the current literature that enables interoperabil- ity and integration between lexical and textual re- sources and the distributional semantics of words, lexical senses and concepts, and collocation prop- erties. By employing the usage of L(L)OD (Linked Linguistic Open Data) technologies, we describe the consolidation of OntoLex-FrAC (Frequency, Attestation, and Corpus Information), an OntoLex- Lemon model that (1) addresses the requirements of corpus-based lexicography (frequency and col- location information) and digital philology (link- ing lexical resources with corpus data), and (2) provides a standard for encoding, storing, and ex- changing vector representations of words along with their lexical concepts, senses, and lemmas. 2 Core Concepts So far, the development of FrAC has been con- ducted in a bottom-up fashion, where uses cases were analyzed and sub-vocabularies for different phenomena have been proposed. This includes fre- quency and attestations (Chiarcos et al., 2020), em- beddings and similarity (Chiarcos et al., 2021) and collocations (Chiarcos et al., 2022). We comple- ment these efforts with a top-down perspective, and we suggest three top-level classes to structure the model as a whole. In addition to that, we provide an OWL2/DL ontology to formalize the vocabu- lary. Restructuring the module entails a number of minor revisions regarding naming and scope of properties and classes, however, we aimed to stay faithful to the original definitions while …

Content truncated to 500 words