Paper Title
Multi-Criteria Recommender System Using Collaborative Filtering and Aspect Based Sentiment Analysis for User Reviews of Food Recipes
Abstract
In the past few years, a significant growth has been seen in the technologies that surround the concept of
Information Filtering. Recommender systems is a subclass of the same and it came into play around 1990 when they were
mentioned as a "digital bookshelf" by Jussi Karlgen in a technical report.
Through this paper, we aim to ideate a working scenario for a Multi-criteria Recommender System (MCRS) that combines
Collaborative Filtering (CF) and Aspect-Based Sentiment Analysis (ABSA) to recommend food recipes based on the aspects
mined from the user reviews.
Our system utilizes CF to identify similar users and recommend recipes based on what their previous likes and preferences
have been. We further incorporate ABSA to comprehend user reviews as it analyzes consumer review data by correlating
sentiments to different aspects and subaspects of the recipe review. These techniques are applied on a dataset that contains
food recipes along with their ratings and textual reviews.
We target to judge the conditions in which the recommender system works best by determining the Mean Average Error
(MAE) in each scenario. Whichever condition produces the least MAE is the ideal scenario for a recommender system to
work in.
Keywords - Information Filtering, Recommendation Systems, Collaborative Filtering, Aspect Based Sentiment Analysis.