Persona or Context? Towards Building Context adaptive Personalized Persuasive Virtual Sales Assistant

Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1035–1047 November 20–23, 2022. ©2022 Association for Computational Linguistics 1035 Persona or Context? Towards Building Context adaptive Personalized Persuasive Virtual Sales Assistant Abhisek Tiwari∗, Sriparna Saha∗, Shubhashis Sengupta†, Anutosh Maitra†, Roshni Ramnani†, Pushpak Bhattacharyya‡ ∗Indian Institute of Technology Patna, India †Accenture Technology Lab, Bengaluru, India ‡Indian Institute of Technology Bombay, India (abhisek_1921cs16, sriparna)@iitp.ac.in Abstract Task-oriented conversational agents are gain- ing immense popularity and success in a wide range of tasks, from flight ticket booking to online shopping. However, the existing sys- tems presume that end-users will always have a pre-determined and servable task goal, which results in dialogue failure in hostile scenarios, such as goal unavailability. On the other hand, human agents accomplish users’ tasks even in a large number of goal unavailability scenar- ios by persuading them towards a very similar and servable goal. Motivated by the limita- tion, we propose and build a novel end-to-end multi-modal persuasive dialogue system incor- porated with a personalized persuasive mod- ule aided goal controller and goal persuader. The goal controller recognizes goal conflict- ing/unavailability scenarios and formulates a new goal, while the goal persuader persuades users using a personalized persuasive strat- egy identified through dialogue context. We also present a novel automatic evaluation met- ric called Persuasiveness Measurement Rate (PMeR) for quantifying the persuasive capa- bility of a conversational agent. The obtained improvements (both quantitative and qualita- tive) firmly establish the superiority and need of the proposed context-guided, personalized persuasive virtual agent over existing tradi- tional task-oriented virtual agents. Further- more, we also curated a multi-modal persua- sive conversational dialogue corpus annotated with intent, slot, sentiment, and dialogue act for e-commerce domain1. 1 Introduction Conversational Artificial Intelligence is gaining popularity and adoption in various fields, owing to its effective task handling and scalability aspects (Xu et al., 2017; Cui et al., 2017; Yan, 2018). In task-oriented dialogue systems, the primary ob- jective of both users and agents is successful task 1Dataset and Code: https://github.com/ NLP-RL/PPMD completion (Chen et al., 2017). Our proposed work is relevant to task-oriented dialogue settings where the proposed agent aims to assist end-users in ac- complishing a task. In real life, when a human sales agent fails to ful- fill consumers’ proposed task requirements, he/she finds a very similar goal and attempts to influence them toward the new goal. Furthermore, end-users prefer to explore and obtain a servable goal by over- looking a little mismatch in many cases. However, existing dialogue systems (Li et al., 2017; Shi and Yu, 2018; Mo et al., 2018; Zhang et al., 2019) ter- minate conversations in such adversarial situations. An illustration has been shown in Figure 1. While the traditional assistant simply terminates the con- versation in the goal unavailability situation, the proposed assistant attempts to serve a very similar phone and persuade the user using context-guided persuasive appeal. Persuasion is a …

Content truncated to 500 words