Near-lossless Binarization of Word Embeddings
This enables efficient NLP model deployment on low-resource devices by significantly compressing embeddings.
The paper tackles the problem of memory and computational inefficiency of real-valued word embeddings by proposing a binarization method that reduces vector size by 97% and speeds up operations by 30 times, with only about a 2% loss in accuracy on NLP tasks.
Word embeddings are commonly used as a starting point in many NLP models to achieve state-of-the-art performances. However, with a large vocabulary and many dimensions, these floating-point representations are expensive both in terms of memory and calculations which makes them unsuitable for use on low-resource devices. The method proposed in this paper transforms real-valued embeddings into binary embeddings while preserving semantic information, requiring only 128 or 256 bits for each vector. This leads to a small memory footprint and fast vector operations. The model is based on an autoencoder architecture, which also allows to reconstruct original vectors from the binary ones. Experimental results on semantic similarity, text classification and sentiment analysis tasks show that the binarization of word embeddings only leads to a loss of ~2% in accuracy while vector size is reduced by 97%. Furthermore, a top-k benchmark demonstrates that using these binary vectors is 30 times faster than using real-valued vectors.