Creation of a Chatbot for video games recommendation with NLP

Abstract

Chatbots are software-based systems designed to interact with humans using text-based natural language. They first appeared in the 1960s, many years passed till the world became ready for their implementation into the real life, as a result of the rapid progress in Natural Language Processing, Machine Learning, and the global presence of text messaging applications. Chatbots are so trendy nowadays, they are being exploited in many aspects, like products recommendation for instance, but their main weakness is that their domain of knowledge is narrow, and it is nearly impossible to create a Chatbot that can tackle multiple domains with today’s knowledge, and thus each Chatbot is created differently to fit the requirements of the specific aspects it is put in. This thesis represents a practical case of a Chatbot deployed for video games recommendation, as supposedly, the field of video games has not been fully exploited yet. The Chatbot uses a combination of traditional Natural Language Processing mixed with other modern techniques such as Neural Networks. The thesis also shows how it is possible to reduce the amount of text data needed to obtain a relatively good intent classification model, by using the semantic similarity of words, or in other words, multi-dimensional meaning representations of words.

Keywords: Conversational Agents, Chatbots, Recommender Systems, Contextmanagement, Machine Learning, Natural Language Processing.