Inducing and Embedding Senses with Scaled Gumbel Softmax
This work addresses the need for interpretable multi-sense embeddings in natural language processing, offering a novel method that balances performance and interpretability, though it is incremental in improving existing approaches.
The authors tackled the problem of learning interpretable word sense embeddings by proposing an unsupervised model that uses a modified Gumbel softmax function for differentiable sense selection, achieving competitive or state-of-the-art results on similarity-based downstream evaluations while also producing interpretable embeddings without duplication.
Methods for learning word sense embeddings represent a single word with multiple sense-specific vectors. These methods should not only produce interpretable sense embeddings, but should also learn how to select which sense to use in a given context. We propose an unsupervised model that learns sense embeddings using a modified Gumbel softmax function, which allows for differentiable discrete sense selection. Our model produces sense embeddings that are competitive (and sometimes state of the art) on multiple similarity based downstream evaluations. However, performance on these downstream evaluations tasks does not correlate with interpretability of sense embeddings, as we discover through an interpretability comparison with competing multi-sense embeddings. While many previous approaches perform well on downstream evaluations, they do not produce interpretable embeddings and learn duplicated sense groups; our method achieves the best of both worlds.