Requirements Elicitation and Response Generation for Conversational Services

Currently, cognitive services provide a more interactive method for understanding users’ requirements via human-machine conversations. In other words, it has to capture users’ requirements from their utterances and respond to them with relevant and suitable service resources. To this end, two phases must be applied: I.sequence planning and real-time detection of user requirements, and II.service resource selection and response generation. The existing works ignore the potential connection between these two phases. To model their connection, the two-phase requirement elicitation method is proposed. In Phase I, this paper proposes a user requirement elicitation framework (URef) to plan a potential requirement sequence grounded on the user profile and personal knowledge base before the conversation. It can also predict user’s true requirement and judge whether the requirement is completed based on the user’s utterance during the conversation. In Phase II, this paper proposes a response generation model based on attention, SaRSNet. It can select the appropriate resource (i.e., knowledge triple) in line with the requirement predicted by URef, and then generates a suitable response for recommendation. The experimental results on the open dataset DuRecDial have been significantly improved compared to the baseline, which proves the effectiveness of the proposed methods. (This Manuscript based partly on work [1] presented at conference (ICWS 2021).)