Generative Adversarial Nets for Multiple Text Corpora
This work addresses the challenge of handling multiple text corpora in NLP, but it appears incremental as it extends existing GAN methods from single to multiple corpora.
The paper tackles the problem of applying generative adversarial nets (GANs) to multiple text corpora, focusing on creating consistent cross-corpus word embeddings and robust bag-of-words document embeddings, and demonstrates that these embeddings improve performance in supervised learning tasks on real-world datasets.
Generative adversarial nets (GANs) have been successfully applied to the artificial generation of image data. In terms of text data, much has been done on the artificial generation of natural language from a single corpus. We consider multiple text corpora as the input data, for which there can be two applications of GANs: (1) the creation of consistent cross-corpus word embeddings given different word embeddings per corpus; (2) the generation of robust bag-of-words document embeddings for each corpora. We demonstrate our GAN models on real-world text data sets from different corpora, and show that embeddings from both models lead to improvements in supervised learning problems.