An Unsupervised, Geometric and Syntax-aware Quantification of Polysemy

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 10565 - 10574 December 7-11, 2022 ©2022 Association for Computational Linguistics An Unsupervised, Geometric and Syntax-aware Quantification of Polysemy Anmol Goel†, Charu Sharma†, Ponnurangam Kumaraguru† †International Institute of Information Technology, Hyderabad anmol.goel@research.iiit.ac.in {charu.sharma, pk.guru}@iiit.ac.in Abstract Polysemy is the phenomenon where a single word form possesses two or more related senses. It is an extremely ubiquitous part of natural language and analyzing it has sparked rich discussions in the linguistics, psychology and philosophy communities alike. With scarce attention paid to polysemy in computational linguistics, and even scarcer attention toward quantifying polysemy, in this paper, we propose a novel, unsupervised framework to compute and estimate polysemy scores for words in multiple languages. We infuse our proposed quantification with syntactic knowledge in the form of dependency structures. This informs the final polysemy scores of the lexicon moti- vated by recent linguistic findings that suggest there is an implicit relation between syntax and ambiguity/polysemy. We adopt a graph based approach by computing the discrete Ollivier Ricci curvature on a graph of the contextual nearest neighbors. We test our framework on curated datasets controlling for different sense distributions of words in 3 typologically diverse languages - English, French and Spanish. The effectiveness of our framework is demonstrated by significant correlations of our quantification with expert human annotated language resources like WordNet. We observe a 0.3 point increase in the correlation coeffi- cient as compared to previous quantification studies in English. Our research leverages contextual language models and syntactic structures to empirically support the widely held theoretical linguistic notion that syntax is intricately linked to ambiguity/polysemy. https://github.com/agoel00/ polysemy 1 Introduction Polysemy is a phenomenon prevalent in everyday language use where the same lexical unit (or word form) is associated with multiple distinct yet re- lated meanings (or senses). Determining which words are polysemous can help in filtering data for linguist studies, creation of sense corpora and the anthropological study of language. Consider the following sentences: 1a His aunt is his legal guardian. 2a The dog would always bark at mailmen. 2b The tree’s bark was rusty brown. 3a The mouth of the wine was dry. 3b I have three mouths to feed. 3c You can see the mouth of the river from here. Polysemy is distinct from monosemy (a word form with only one meaning; 1a) and homonymy (multiple unrelated senses of the same word form; 2a-b). The polysemous senses of a word often have metonymic (3a-b) or metaphorical (3c) relations among them (Vicente and Falkum, 2017). Poly- semy is a central feature of natural languages and proliferates almost every word to varying degrees in the lexicon of a language. Attempts (Pianta- dosi et al., 2012) at explaining the presence of am- biguity1 in language suggest that polysemy is a desirable property for language systems since it al- lows efficient communication by allowing simpler units to be reused. Ambiguity and polysemy have sparked debate among linguists and philosophers for decades but relatively little …

Content truncated to 500 words