SeqDiffuSeq: Text Diffusion with Encoder-Decoder Transformers
This addresses the problem of extending diffusion models to natural language generation for NLP researchers, representing an incremental advancement in adapting image-based techniques to text.
The paper tackles the challenge of applying continuous diffusion models to discrete text for sequence-to-sequence generation by proposing SeqDiffuSeq, which uses an encoder-decoder Transformer architecture with self-conditioning and an adaptive noise schedule, resulting in improved text quality and inference time.
Diffusion model, a new generative modelling paradigm, has achieved great success in image, audio, and video generation. However, considering the discrete categorical nature of text, it is not trivial to extend continuous diffusion models to natural language, and text diffusion models are less studied. Sequence-to-sequence text generation is one of the essential natural language processing topics. In this work, we apply diffusion models to approach sequence-to-sequence text generation, and explore whether the superiority generation performance of diffusion model can transfer to natural language domain. We propose SeqDiffuSeq, a text diffusion model for sequence-to-sequence generation. SeqDiffuSeq uses an encoder-decoder Transformers architecture to model denoising function. In order to improve generation quality, SeqDiffuSeq combines the self-conditioning technique and a newly proposed adaptive noise schedule technique. The adaptive noise schedule has the difficulty of denoising evenly distributed across time steps, and considers exclusive noise schedules for tokens at different positional order. Experiment results illustrate the good performance on sequence-to-sequence generation in terms of text quality and inference time.