A neural topic model with word vectors and entity vectors for short texts

作者：

Highlights：

• We first propose an unsupervised topic model named VAETM for short texts.

• We then consider label information in dataset to boost VAETM.

• A KL-divergence-based algorithm is used to infer approximate posterior distribution.

• Extensive experiments demonstrate our models outperform state-of-the-art baselines.

摘要

•We first propose an unsupervised topic model named VAETM for short texts.•We then consider label information in dataset to boost VAETM.•A KL-divergence-based algorithm is used to infer approximate posterior distribution.•Extensive experiments demonstrate our models outperform state-of-the-art baselines.

论文关键词：Topic model,Short text,Variational auto-encoder,Word embedding,Entity embedding

论文评审过程：Received 7 February 2020, Revised 10 November 2020, Accepted 25 November 2020, Available online 11 December 2020, Version of Record 11 December 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2020.102455