کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

A novel approach for recommender systems based on query likelihood and sentiment analysis

نویسندگان :

Mohammadreza Soltaninezhad¹ Alireza Basiri²

1- دانشگاه صنعتی اصفهان 2- دانشگاه صنعتی اصفهان

کلمات کلیدی :

Recommender System،Collaborative Filtering،Query Likelihood Model،Sentiment Analysis

چکیده :

Collaborative Filtering, a widely used approach in recommendation systems, frequently encounters significant limitations, including its incapacity to analyze textual material, sensitivity to sparse datasets, and challenges in understanding complex user preferences. To address these challenges, this paper introduces a novel methodology that leverages Query Likelihood Models and Sentiment Analysis to enhance recommendation accuracy. Query likelihood provides a probabilistic framework for modeling user preferences as queries and item reviews as documents, offering a way to combine textual data in the recommendation process. Sentiment analysis improves performance by quantifying the sentiment of review texts, weighting key words based on their significance to user contentment. Sentiment scores are obtained using a Bi-LSTM sentiment analysis model and incorporated into the QLM framework. The system considers sentiment-weighted keywords extracted from user reviews as user queries and builds item documents by aggregating their related reviews. These sentiment scores assist in modifying the log-likelihood calculations, which are then normalized and scaled to predict final ratings. We use the Amazon Food Reviews dataset for evaluations; our experiment results indicate that the novel proposed model, which is a combination of the query likelihood model with sentiment-weighted keywords and traditional collaborative filtering, outperformed the traditional collaborative filtering model across nearly all evaluation metrics. Thus, the proposed method offers a scalable and interpretable framework for improving the accuracy of Recommendation Systems.