Conversational AI - A Retrieval Based Chatbot

EasyChair Preprint 4020

Abstract

Conversational AI can be simply defined as human-computer interaction through natural conversations. This may be through a chatbot on a website or any social messaging app, a voice assistant or any other interactive messaging-enabled interfaces. This system will allow people to ask queries, get opinions or recommendations, execute needed transactions, find support or otherwise achieve a goal through conversations. Chatbots are basically online human-computer dialog system with natural language. Currently, advancements in natural language processing and machine learning mechanism have improved chatbot technology. More commercial and social media platforms are now employing this technology in their services. Organisations demands artificial intelligence based improvements in chatbot adoption and thus it became one of the hot research. In this work, a task-oriented retrieval based chatbot has been proposed on a bus ticket booking domain which is built using a Deep Neural Network. The sequence of questions that are being asked by multiple users with different kinds of personas are taken as the input for the system. And accordingly the retrieval based system produces meaningful responses. The generated responses are evaluated manually. The results show that the generated answers are meaningful in most of the cases.

Keyphrases: Chatbots, Retrieval Based Model, deep learning, neural network

@booklet{EasyChair:4020,
  author    = {Aiswarya Surendran and Reena Murali and Remesh K. R. Babu},
  title     = {Conversational AI - A Retrieval Based Chatbot},
  howpublished = {EasyChair Preprint 4020},
  year      = {EasyChair, 2020}}