Opinion-based intelligent recommender system

• Main Author: Poh, Ying Xuan
• Other Authors: Li Fang
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/147996
• Institution: Nanyang Technological University

With the recent development of Natural Language Processing (NLP), it is possible to extract sentiments from a text with given aspects. Collaborative Filtering techniques are used to recommend items to generate personalised recommendations based on similar users' preferences. Deep learning has grown popular in recent years for its immense accuracy over massive datasets. In this paper, we proposed to design an opinion-based intelligent recommender system utilising deep learning. This system incorporates aspect-based sentiment analysis to understand and quantify text, followed by performing collaborative filtering techniques to build a recommender system. For the aspect-based sentiment analysis task, it is executed by converting texts sentences into auxiliary sentences followed by classification training using Bidirectional Encoder Representations from Transformers(BERT) to quantify texts into ratings. For collaborative filtering, it is accomplished using a modified Neural Collaborative Filtering(NCF) that learns the user-item interactions by recognising the relationship between aspects and ratings to provide recommendations to different users. The results are evaluated towards the end and could be used for real-life applications.