Kernelized Bayesian Softmax for Text Generation
This addresses the challenge of handling multiple word senses in text generation for NLP applications, representing an incremental improvement over existing embedding methods.
The paper tackles the problem of ambiguous word senses in neural text generation by proposing KerBS, a method that uses Bayesian composition and learned kernels to improve token embeddings, resulting in significant performance boosts across several tasks.
Neural models for text generation require a softmax layer with proper token embeddings during the decoding phase. Most existing approaches adopt single point embedding for each token. However, a word may have multiple senses according to different context, some of which might be distinct. In this paper, we propose KerBS, a novel approach for learning better embeddings for text generation. KerBS embodies two advantages: (a) it employs a Bayesian composition of embeddings for words with multiple senses; (b) it is adaptive to semantic variances of words and robust to rare sentence context by imposing learned kernels to capture the closeness of words (senses) in the embedding space. Empirical studies show that KerBS significantly boosts the performance of several text generation tasks.