eduzhai > Applied Sciences > Engineering >

Speech Sentiment and Customer Satisfaction Estimation in Socialbot Conversations

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: For an interactive agent, such as task-oriented spoken dialog systems orchatbots, measuring and adapting to Customer Satisfaction (CSAT) is critical inorder to understand user perception of an agent s behavior and increase userengagement and retention. However, an agent often relies on explicit customerfeedback for measuring CSAT. Such explicit feedback may result in potentialdistraction to users and it can be challenging to capture continuously changinguser s satisfaction. To address this challenge, we present a new approach toautomatically estimate CSAT using acoustic and lexical information in the AlexaPrize Socialbot data. We first explore the relationship between CSAT andsentiment scores at both the utterance and conversation level. We theninvestigate static and temporal modeling methods that use estimated sentimentscores as a mid-level representation. The results show that the sentimentscores, particularly valence and satisfaction, are correlated with CSAT. Wealso demonstrate that our proposed temporal modeling approach for estimatingCSAT achieves competitive performance, relative to static baselines as well ashuman performance. This work provides insights into open domain socialconversations between real users and socialbots, and the use of both acousticand lexical information for understanding the relationship between CSAT andsentiment scores.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×