Avis sur les restaurants en ligne: détection et analyse quantitative

Résumé

The study examines, from various perspectives, restaurant evaluations written by customers and published online. From a manual observation of the corpus, we propose a model to study evaluations of experiences in restaurants; we show that evaluations can have several functions and contain different types of information: reviewers write reviews to give opinions, make suggestions, express intentions, and describe their experiences. We conduct various quantitative analyses using corpus linguistics methods which offer the possibility to identify the keywords of the reviews as well as their grammatical and topologocial characteristics, including keyword analysis, the distribution of parts of speech and the positions of evaluation categories within reviews. We finally try to automatically detect the evaluation categories in our corpus. We test and validate a supervised learning approach based on our model completed with other linguistic and textual features.