Automatic Taxonomy Expansion in IndoWordNet: WordNet of Indian Languages

Presenter

Bornali Phukon

Date

23-Oct

Abstract

In this talk, I will discuss our research on automatic taxonomy expansion within WordNet, with a specific focus on IndoWordNet, which pertains to Indian languages. Our research strongly emphasizes the profound significance of this work for the realm of Natural Language Processing (NLP), especially in addressing the persistent challenge of limited taxonomy coverage, particularly in low-resource language contexts. We thoroughly investigate the issue of missing relations within WordNet and introduce the innovative “TEAM” and “LG-TEAM” models. These models function within a multi-task learning framework to proficiently manage the intricate “Attach” and “Merge” operations, unifying them for seamless taxonomy expansion. Through our research, we aim to optimize taxonomy expansion, offering the potential for improved precision and effectiveness in NLP applications, with a special emphasis on supporting linguistic communities that are often underrepresented. I invite you to join me to explore the impact of this research in the NLP field and its potential implications for automatic speech recognition in under-resourced languages, together with me.