Improving Disentangled Text Representation Learning with Information-Theoretic Guidance
This addresses a key problem in NLP for applications such as personalized dialogue systems, though it is an incremental improvement over existing disentanglement methods adapted from other data types.
The paper tackles the challenge of learning disentangled text representations for NLP tasks like conditional text generation and style transfer by proposing an information-theoretic method that minimizes a mutual information upper bound to separate style and content embeddings. Experiments show high-quality results in content and style preservation.
Learning disentangled representations of natural language is essential for many NLP tasks, e.g., conditional text generation, style transfer, personalized dialogue systems, etc. Similar problems have been studied extensively for other forms of data, such as images and videos. However, the discrete nature of natural language makes the disentangling of textual representations more challenging (e.g., the manipulation over the data space cannot be easily achieved). Inspired by information theory, we propose a novel method that effectively manifests disentangled representations of text, without any supervision on semantics. A new mutual information upper bound is derived and leveraged to measure dependence between style and content. By minimizing this upper bound, the proposed method induces style and content embeddings into two independent low-dimensional spaces. Experiments on both conditional text generation and text-style transfer demonstrate the high quality of our disentangled representation in terms of content and style preservation.