Naslov
Syntactic Dependency Networks: Cognitive Aspects of Hierarhical Multi-Layer Structures

Autori
Ban Kirigin, Tajana ; Bujačić Babić, Sanda ; Perak, Benedikt

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Book of Abstracts LAP 2022 / - Dubrovnik, 2022, 60-61

Skup
Logic and Applications 2022

Mjesto i datum
Dubrovnik, Hrvatska, 26.09.2022. - 29.09.2022

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Universal dependencies ; Dependency graph analysis ; Semantics ; Syntax

Sažetak
Lexical items, i.e. words are building blocks of linguistic structures such as phrases, sentences, paragraphs, texts, etc that encode the conceptual content perceived, or imagined, by human cognitive processing. The sequence of lexical items within a linguistic structure is organised by a set of syntactic relations construing a conceptual relation and representing an emergent semantic structure or a meaning. According to the Universal Dependencies (UD), a framework for morphosyntactic annotation of human language, which to date has been used to create treebanks for more than 100 languages (De Marneffe et al. 2022), the classification of syntactic relations offers a linguistic representation that is useful for morphosyntactic research, semantic interpretation, and for practical natural language processing across different human languages. A syntactic dependency, in general, is a binary phrasal asymmetric grammar relation with a lexical head and other lexical items as dependents of that head, represented in diagrams by an arrow from the head word to the dependent word. This work will present the types of UD dependency relations and their potential for multi-layered syntacticsemantic analysis of texts. The types of dependencies are classified according to the emergent cognitive aspects of the syntactic- semantic complexity represented in terms of the hierarchical emergent ontological schema of semantic roles and agent-based representation. The multi-layered formalization of the dependency structures extracted from tagged corpora can be used for a graph representation of common knowledge of conceptual entities, attributes and processes as well as downstream NLP applications, such as lexical labelling, figurative speech identification and dictionary-based sentiment analysis systems.

Izvorni jezik
Engleski

Znanstvena područja
Matematika, Računarstvo, Filozofija

POVEZANOST RADA