Search papers, labs, and topics across Lattice.
This paper investigates the effectiveness of various word sense induction (WSI) methods, including LLM-based approaches, in the absence of sense-annotated data. The authors highlight methodological issues in current WSI evaluation and propose a SemCor-derived dataset that respects original corpus polysemy and frequency distributions. Their experiments reveal that no unsupervised method, including their LLM-based approach, outperforms a simple "one cluster per lemma" heuristic, although data augmentation and leveraging Wiktionary can improve performance.
Despite the rise of LLMs, unsupervised word sense induction remains an open problem, with even advanced models failing to beat a naive baseline.
In the absence of sense-annotated data, word sense induction (WSI) is a compelling alternative to word sense disambiguation, particularly in low-resource or domain-specific settings. In this paper, we emphasize methodological problems in current WSI evaluation. We propose an evaluation on a SemCor-derived dataset, respecting the original corpus polysemy and frequency distributions. We assess pre-trained embeddings and clustering algorithms across parts of speech, and propose and evaluate an LLM-based WSI method for English. We evaluate data augmentation sources (LLM-generated, corpus and lexicon), and semi-supervised scenarios using Wiktionary for data augmentation, must-link constraints, number of clusters per lemma. We find that no unsupervised method (whether ours or previous) surpasses the strong"one cluster per lemma"heuristic (1cpl). We also show that (i) results and best systems may vary across POS, (ii) LLMs have troubles performing this task, (iii) data augmentation is beneficial and (iv) capitalizing on Wiktionary does help. It surpasses previous SOTA system on our test set by 3.3\%. WSI is not solved, and calls for a better articulation of lexicons and LLMs'lexical semantics capabilities.
