Fine-grained sentiment analysis on food recipe reviews

Abstract

Among an oversized variety of food recipes available on the web, only some recipes are posted appropriately. The dish might not appear identical when preparing the dish by reading a specific food direction. It is imperative to find a suitable recipe to prepare a meal properly. Customer satisfaction is considered a good measure to find a suitable recipe. Therefore, recipes should have wise ratings from users. The ratings and comments facilitate users to search out suitable recipes. Sentiments, analysis techniques and lassifier algorithms will be used to classify sentiments based primarily on words. Sentiment classifiers are used for binary classification (positive or negative). However, in this analysis, five separate categories of sentiments are used. They are strongly negative, weakly negative, neutral, weakly positive and strongly positive. This analysis used a supervised learning technique that categorized each new document by one or more class labels from a set of predefined classes. An improved multinomial naïve Bayes algorithmic program is employed to develop the model and compare the model accuracy and precision with fine-grained and binary classification. In this work, an algorithm was developed to choose the suitable recipe by analyzing user ratings.