Document-based topic coherence measures for news media text

作者：

Highlights：

• Novel class of document-based coherence measures for news topics proposed and evaluated.

• High-performing document-based coherence measure identified.

• Document-based coherence measures contrasted with state-of-art word-based measures.

• Application of document-based measures for semi-automated topic discovery.

摘要

•Novel class of document-based coherence measures for news topics proposed and evaluated.•High-performing document-based coherence measure identified.•Document-based coherence measures contrasted with state-of-art word-based measures.•Application of document-based measures for semi-automated topic discovery.

论文关键词：Topic models,Topic coherence,Topic model evaluation,Text analysis,News text,Exploratory analysis

论文评审过程：Received 20 April 2018, Revised 7 July 2018, Accepted 29 July 2018, Available online 30 July 2018, Version of Record 7 August 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.07.063