Combating the Curse of Multilinguality in Cross-Lingual WSD by Aligning Sparse Contextualized Word Representations
This addresses the curse of multilinguality in cross-lingual WSD for NLP applications, representing an incremental improvement with specific gains.
The paper tackled cross-lingual word sense disambiguation by using large pre-trained monolingual models with a contextualized mapping mechanism and sparse representations, resulting in a 6.5-point increase in average F-score from 62.0 to 68.5 across 17 languages.
In this paper, we advocate for using large pre-trained monolingual language models in cross lingual zero-shot word sense disambiguation (WSD) coupled with a contextualized mapping mechanism. We also report rigorous experiments that illustrate the effectiveness of employing sparse contextualized word representations obtained via a dictionary learning procedure. Our experimental results demonstrate that the above modifications yield a significant improvement of nearly 6.5 points of increase in the average F-score (from 62.0 to 68.5) over a collection of 17 typologically diverse set of target languages. We release our source code for replicating our experiments at https://github.com/begab/sparsity_makes_sense.