Home /Research /PESUBot: An Empathetic Goal Oriented Chatbot
LEARNING

PESUBot: An Empathetic Goal Oriented Chatbot

Hemkesh Vijaya Kumar, Jeevan Nagaraj, Mohammed Irfan, Nilisha Maheshwari, Pooja Balusani, Pratik Chatterjee, Gowri Srinivasa

Year
2018
Citations
7

Abstract

`PESUBot', a robot-receptionist, is a chatbot that is designed to be deployed in university premises to answer questions related to the university and also maintains a general conversation with its user. The chatbot was built specifically for PES University, Bangalore. The dataset used for building this chatbot was PES University specific data which was scraped from the university's website, personality data of the chatbot which was custom formulated and the general conversation dataset was curated from reddit data. The chatbot is composed of 2 prominent models, one of which is a retrieval-based model that can answer questions about the university and about the chatbot itself while the other is a generative model that answers any other general question that is passed on to it by the first model. Both these models are built using Recurrent Neural Networks. The first model had an average per response accuracy of 92% while the second model had an average BLEU score of 0.53.

Keywords

ChatbotConversationComputer scienceArtificial intelligenceWorld Wide WebNatural language processingInformation retrievalHuman–computer interactionPsychology

Related papers

Browse all LEARNING papers