Making sense of distributional semantics
Year: 2023
Type: book-chapter
Abstract: This chapter focuses on the ‘semantics’ in ‘distributional semantics’: what do clusters in distributional models look like and what do they mean? Based on analyses performed on 32 Dutch nouns, adjectives, and verbs from a synchronic, semasiological perspective, the chapter shows that there is no parameter setting for distributional models that returns an optimal solution across the board and second, that the token clouds, that is, the clusters within the models, do not systematically match senses as one would find them in dictionaries. Token clusters in distributional models may take different shapes, depending on the frequency and distinctiveness of the context words, and such clusters do not coincide unambiguously with lexicographic senses.
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4388932315