Evolutionary Generative Adversarial Networks for Sentence Generation from Keyword
Ray Oshikawa, Kotaro Pindur, Hitoshi Iba · 2022 IEEE World Conference on Applied Intelligence and Computing (AIC) · 2022
Natural language processing has been recognized as an important goal for computers. Especially, generating sentence as we needed is very practical goal. However, the current GAN implementations for sentence generation ignore context. Generating sentences based on a simple corpus by specifying keywords will bring us one step closer to achieving generating along context. This research has made the following two contributions. First, we confirmed the effectiveness of genetic programming in keyphrase selection, built a non-black-box keyphrase selection model. Secondly, we improved the evolutionary GAN and realized sentence generation based on keywords. There is little integration of natural language processing by evolutionary computation methods, so this work is a good step.