Reconstruction of Word Embeddings from Sub-Word Parameters
This addresses the trade-off between performance and model size for NLP practitioners, though it appears incremental.
The paper tackles the problem of model size increase from pre-trained word embeddings by reconstructing them from sub-word parameters, achieving interesting results on word similarity, analogy, and part-of-speech tagging tasks.
Pre-trained word embeddings improve the performance of a neural model at the cost of increasing the model size. We propose to benefit from this resource without paying the cost by operating strictly at the sub-lexical level. Our approach is quite simple: before task-specific training, we first optimize sub-word parameters to reconstruct pre-trained word embeddings using various distance measures. We report interesting results on a variety of tasks: word similarity, word analogy, and part-of-speech tagging.