CONTEXT-AWARE RECOMMENDATION SYSTEM IN ECOMMERCE USING A DOMAIN SPECIFIC CHATBOT

Abstract
Recommendation systems are used in almost every e-commerce platform. With the amount of information increasing every day with almost unlimited products available, users tend to find it difficult to make a decision on these products. This research aims to develop a context-aware recommendation system that communicates to the user through a domain specific chatbot. The chatbot converses in three languages, English, Yoruba and Igbo. In this research, language is brought in as context to address existing gaps in literature related to e-commerce recommendation system. The developed system contains three main components they include, collaborative filtering, content based recommendation system and a chatbot. The collaborative and content-based recommendation systems uses user ratings and genre as the input to the cosine similarity and tf-idf functions. The chatbot was trained using a Keras sequential API, which created a deep neural network, and an interface was designed using python library tkinter and customtkinter. The chatbot uses a json file that contains the intents, and the recommendation systems use a modified 100k movielens dataset that contains Hollywood and Nollywood movies. The developed system is able to contribute to existing body of knowledge by generating an indigenous language corpus that can be used and modified by future researchs related to conversational system particularly in ecommerce recommendation. This research, will enable ecommerce platforms improve customer satisfaction, which would lead to improved revenue. In the aspect of trust, users will be able to have conversation with the system through the chatbot, which can increase trust between the users and the platform.
Description
Keywords
Citation