Contextual Topic Modeling For Dialog Systems

  • 2018-10-18 16:19:16
  • Chandra Khatri, Rahul Goel, Behnam Hedayatni, Angeliki Metanillou, Anushree Venkatesh, Raefer Gabriel, Arindam Mandal
  • 11

Abstract

Accurate prediction of conversation topics can be a valuable signal forcreating coherent and engaging dialog systems. In this work, we focus oncontext-aware topic classification methods for identifying topics in free-formhuman-chatbot dialogs. We extend previous work on neural topic classificationand unsupervised topic keyword detection by incorporating conversationalcontext and dialog act features. On annotated data, we show that incorporatingcontext and dialog acts leads to relative gains in topic classificationaccuracy by 35% and on unsupervised keyword detection recall by 11% forconversational interactions where topics frequently span multiple utterances.We show that topical metrics such as topical depth is highly correlated withdialog evaluation metrics such as coherence and engagement implying thatconversational topic models can predict user satisfaction. Our work fordetecting conversation topics and keywords can be used to guide chatbotstowards coherent dialog.

 

Quick Read (beta)

loading the full paper ...