Empirical Antonym Implementation in the UMLS SPECIALIST Lexicon

Area of Science:

• Computational Linguistics
• Lexicography
• Natural Language Processing (NLP)

Background:

• Antonyms, words with opposite meanings, are vital for NLP tasks.
• Existing antonym resources may lack comprehensive features or broad coverage.
• The need for systematic antonym generation for NLP applications is recognized.

Purpose of the Study:

• To systematically generate a large set of canonical antonyms.
• To develop and utilize multiple source models for antonym discovery.
• To release a comprehensive lexicon of antonyms with associated features for NLP.

Main Methods:

• Development of five distinct source models for antonym generation.
• Utilizing negation rules, derivational morphology, corpus co-occurrences, and semantic networks.
• Systematic generation and canonicalization of antonym pairs and their features.

Main Results:

• Generation of 13,000 canonical antonyms.
• Inclusion of features such as bounded types, canonical domains, and negations.
• Release of the SPECIALIST Lexicon 2025, offering broad antonym coverage.

Conclusions:

• The developed systematic approach effectively generates a large, feature-rich set of canonical antonyms.
• The SPECIALIST Lexicon 2025 provides valuable resources for NLP research and applications.
• Analysis of antonym sources, canonicity, and features offers insights into lexical relationships.