Process models of interrelated speech intentions from online health-related conversations

Highlights：

• An automatic approach to reveal process models of interrelated speech intentions from conversational turns is proposed.

• A domain-independent speech intentions taxonomy is created and a validated tagged corpus of Reddit discussions is released.

• Supervised classifiers for speech intentions discovery are trained and a method to transform conversations in well-defined, representative logs of verbal behavior for process mining exploration is proposed.

• The obtained results are promising and the extracted knowledge opens new perspectives on understanding beliefs’ and behavioral intentions’ formation in and from speech.

摘要

•An automatic approach to reveal process models of interrelated speech intentions from conversational turns is proposed.•A domain-independent speech intentions taxonomy is created and a validated tagged corpus of Reddit discussions is released.•Supervised classifiers for speech intentions discovery are trained and a method to transform conversations in well-defined, representative logs of verbal behavior for process mining exploration is proposed.•The obtained results are promising and the extracted knowledge opens new perspectives on understanding beliefs’ and behavioral intentions’ formation in and from speech.

论文关键词：Intention mining,Text mining,Natural language processing,Machine learning,Process mining,Speech acts,Speech intentions,Conversational processes,Conversation analysis

论文评审过程：Received 5 November 2017, Revised 25 June 2018, Accepted 28 June 2018, Available online 18 July 2018, Version of Record 6 November 2018.