Improving Big Data Recommendation System Performance using NLP techniques with multi attributes

• Published in: Informatica (Ljubljana) 2024-02, Vol.48 (5), p.63-70
• Main Authors: K. Omar, Hoger, Frikha, Mondher, Jumaa, Alaa Khalil
• Format: Article
• Language: English
• Subjects: Algorithms; Big Data; Data analysis; Datasets; Machine learning; Natural language processing; Ratings; Ratings & rankings; Recommender systems; Social networks; System effectiveness
• ISSN: 0350-5596; 1854-3871
• DOI: 10.31449/inf.v48i5.5255

Due to the wide availability of big data, institutions and companies are currently concentrating on developing highly effective recommender systems for their users. Traditional recommender systems use standard information such as user, item, and ratings. However, this data may not be suffi cient for precise results. To enhance accuracy, it is recommended to include additional information such as textual data in the recommendation system. When dealing with large textual data, employing Natural Language Processing (NLP) techniques is essential for effective data analysis. Therefore, this work proposed a novel big data recommender system that enhances collaborative filtering (CF) results by leveraging NLP techniques and dealing with multiple attributes. The study constructs two big data recommendation system models using a machine learning algorithm. In both models, the Alternating Least Squares (ALS) algorithm within the Apache Spark big data tool has been used. The first model did not incorporate NLP techniques, while the second model considered the novel NLP techniques by taking into account the user's review comments. A dataset of more than 3 million ratings and reviews was gathered from the Amazon website with a size of 3.1 GB. The results showed significant improvement after incorporating the suggested NLP-based techniques with multiple attributes.