ATM: Adversarial-neural Topic Model

作者：

Highlights：

• To our best knowledge, the Adversarial-neural Topic Model (ATM), proposed in this work, is the first attempt to model topics using adversarial training.

• The proposed ATM could generate more coherent topics (considering five coherence measures) compared with the state-of-the-art baselines.

• In practical, the proposed ATM has good portability for other NLP task with limited modification of the model.

摘要

•To our best knowledge, the Adversarial-neural Topic Model (ATM), proposed in this work, is the first attempt to model topics using adversarial training.•The proposed ATM could generate more coherent topics (considering five coherence measures) compared with the state-of-the-art baselines.•In practical, the proposed ATM has good portability for other NLP task with limited modification of the model.

论文关键词：Generative adversarial net,Neural-based topic model,Open domain event extraction,Topic modeling

论文评审过程：Received 15 January 2019, Revised 6 July 2019, Accepted 7 August 2019, Available online 17 August 2019, Version of Record 17 August 2019.

论文官网地址：https://doi.org/10.1016/j.ipm.2019.102098