Recipe Recommendation System (Rec-Res) Using Tf-Idf And Doc2vec

Abstract

With the growing necessity for personalized dish suggestions, recipe recommendation systems are increasingly popular nowadays. In this paper, a new recipe recommendation system is presented that integrates two state-of-the-art natural language processing (NLP) methodologies, TF-IDF (Term Frequency-Inverse Document Frequency) and Doc2Vec, in order to come up with personalized recipe suggestions considering user preferences as well as dish descriptions. The system initially employs TF-IDF to retrieve significant ingredients steps, and categories from the recipe corpus, mentioning the significant terms in the recipe vocabulary. Later, Doc2Vec is used to transform the recipe text into vector representations, allowing the system to understand the semantic connections between recipes. By comparing user preferences against the recipe vectors, the system produces personalized and contextually relevant recommendations. Experimental outcomes emphasize that the suggested method vastly enhances recommendation correctness and user experience compared to classic keyword-based suggestion systems. The paper gives an important contribution by introducing a wiser, more scalable, and user-focused recipe suggestion system relying on state-of-the-art NLP methods.