학술논문
Visual analytics for token-based distributional semantics
Document Type
Chapter
Author
Geeraerts, Dirk, author; Speelman, Dirk, author; Heylen, Kris, author; Montes, Mariana, author; De Pascale, Stefano, author; Franco, Karlien, author; Lang, Michael, author
Source
Lexical Variation and Change : A Distributional Semantic Approach, 2023.
Subject
Language
English
Abstract
The modelling workflow described in Chapter 3 produces token-level distance matrices: one matrix per model, each indicating the pairwise dissimilarity between the occurrences of a certain word in a sample, according to that model. However, because of the large number of tokens in the sample and the diversity of models produced by multiple parameters, such output is challenging to interpret. In this chapter we will describe the steps followed to process the distance matrices and obtain a more manageable format, as well as a visual analytics tool designed to explore the results. The chapter introduces a number of notions that are relevant for the way in which the models discussed in Chapter 5 are analysed and how the token-level plots in Chapters 6, 9, and 10 can be interpreted.