Semantic Space of Parts of Speech

2026-08-15

AI Summary

This study addresses the limitations of traditional discrete part-of-speech (POS) categorization in capturing semantic and syntactic ambiguity by proposing a fuzzy POS continuum perspective. Integrating word2vec, neural network-based dimensionality reduction, and the Universal Dependencies framework, we construct a three-dimensional semantic space to quantify POS boundaries and prototype distributions. Successfully mapping thousands of lexical items across five languages into this space, the research visually reveals clustering characteristics of POS prototypes, fuzzy boundaries, and their semantic interrelations. By transcending conventional discrete classification paradigms, this work provides a novel visualization framework and theoretical foundation for understanding the cross-linguistic nature of parts of speech, offering significant insights into the gradient properties of grammatical categories.

Abstract

Parts of speech categorization is understood in the European linguistic tradition as crisp categorization, which is also reflected in corpus linguistics, where each disambiguated token is assigned exactly one POS. However, the assigned categories are largely determined by arbitrary decisions distilled into annotation manuals. Since some words stand between parts of speech in their semantics or typical syntax, and some parts of speech are closer to each other than others, POS categorization seems inherently fuzzy. We analyze this fuzziness using word2vec embeddings, training a neural network to reduce their high dimensionality to three dimensions relevant for determining parts of speech. This creates a three-dimensional space onto which we map several thousand words, revealing which are prototypical and which lie on the boundaries, and visualizing relationships between parts of speech. The study uses Universal Dependencies POS tags for French, Czech, Finnish, Russian, and English.

Problem

• Parts of Speech
• Fuzzy Categorization
• Semantic Space
• Word Embeddings

Innovation

• POS fuzziness
• word2vec embeddings
• dimensionality reduction
• semantic space visualization
• Universal Dependencies