MTSI-BERT: A Session-aware Knowledge-based Conversational Agent

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 717–725 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 717 MTSI-BERT: A Session-aware Knowledge-based Conversational Agent Matteo A. Senese, Giuseppe Rizzo, Mauro Dragoni, Maurizio Morisio LINKS Foundation, Fondazione Bruno Kessler, Politecnico di Torino matteo.senese@linksfoundation.com, giuseppe.rizzo@linksfoundation.com, dragoni@fbk.eu, maurizio.morisio@polito.it Abstract In the last years, the state of the art of NLP research has made a huge step forward. Since the release of ELMo (Peters et al., 2018), a new race for the leading scoreboards of all the main linguistic tasks has begun. Several models have been published achieving promising results in all the major NLP applications, from question answering to text classification, passing through named entity recognition. These great research discoveries coincide with an increasing trend for voice-based technologies in the customer care market. One of the next biggest challenges in this scenario will be the handling of multi-turn conversations, a type of conversations that differs from single-turn by the presence of multiple related interactions. The proposed work is an attempt to exploit one of these new milestones to handle multi-turn conversations. MTSI-BERT is a BERT-based model achieving promising results in intent classification, knowledge base action prediction and end of dialogue session detection, to determine the right moment to fulfill the user request. The study about the realization of PuffBot, an intelligent chatbot to support and monitor people suffering from asthma, shows how this type of technique could be an important piece in the development of future chatbots. Keywords: Multi-turn, Conversational AI, Goal-Oriented, BERT 1. Introduction Conversational AI is the field of Artificial Intelligence that investigates the techniques to allow machines to interact with us using natural language conversations. Such a task is among the most challenging of the entire NLP field and also one of the hottest topics in nowadays customer care market. The possibility to automate customer care process using an intelligent agent that can understand the user request, assist him and help him to fulfill the final goal is a great desire of companies all around the world. Nevertheless, Conversa- tional AI can also be exploited in medical scenario to help doctors in monitoring the status of their patients and avoid the latter to make long lines at the hospital for trivial sick- nesses that do not require the expert intervention. Today the big tech players such as Google and Amazon have al- ready deployed conversational agents capable to assist the user in the fulfillment of daily tasks through the use of voice (Tulshan and Dhage, 2019). Such agents can schedule ap- pointments, control IoT devices in a room, play music and perform searches on the Web. This type of agents simulates what a typical assistant can do, thus taking the name of vir- tual assistants. Virtual assistants like the Google Assistant and the Amazon Alexa are today widely used in everyday life in an imperative setup, namely a person asks once and the assistant replies and acts right after (“turn …

Content truncated to 500 words