An Empirical Examination of Online Restaurant Reviews

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 4942–4947 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 4942 An Empirical Examination of Online Restaurant Reviews Hyun Jung KANG, Iris ESHKOL-TARAVELLA MoDyCo UMR7114, 200 Avenue de la R´epublique, 92001 Nanterre, France {hyunjung.kang, ieshkolt}@parisnanterre.fr Abstract In the wake of (Pang et al., 2002; Turney, 2002; Liu, 2012) inter alia, opinion mining and sentiment analysis have focused on extracting either positive or negative opinions from texts and determining the targets of these opinions. In this study, we go beyond the coarse-grained positive vs. negative opposition and propose a corpus-based scheme that detects evaluative language at a finer-grained level. We classify each sentence into one of four evaluation types based on the proposed scheme: (1) the reviewer’s opinion on the restaurant (positive, negative, or mixed); (2) the reviewer’s input/feedback to potential customers and restaurant owners (suggestion, advice, or warning) (3) whether the reviewer wants to return to the restaurant (intention); (4) the factual statement about the experience (description). We apply classical machine learning and deep learning methods to show the effectiveness of our scheme. We also interpret the performances that we obtained for each category by taking into account the specificities of the corpus treated. Keywords: Opinion Mining, Online Restaurant Reviews, Machine Learning 1. Introduction As an increasing number of people are consulting online reviews in their decision-making process, online reviews have become a precious asset in various disciplines such as marketing, information science and linguistics. Eval- uation encodes the reviewers’ point of view and provides subjective information based on individual preferences and tastes. With the vast amount of data in the digital age, the state-of-the-art technology alone cannot provide user’s nu- anced understanding. It is with a linguistic approach that we can uncover new associations between language use, social bonds, and non-linguistic aspects of communication. In this context, classifying an evaluative text as positive or negatives–as it is the case in most practical applications–is not sufficient for opinion mining (or sentiment analysis) ap- plications. In this study, we go further to fully understand online reviews by encompassing users’ needs and predict- ing their future actions. The main contributions of this work are: • introducing an expanded classification scheme for evaluative language beyond the simple distinction of positive and negative; • studying online reviews from a linguistic perspective; • demonstrating the effectiveness of our scheme by con- ducting the experiments. 2. Related Work While there is much literature regarding opinion mining, there are fewer works on suggestion and intention mining. We will briefly describe related works in the areas of opinion, suggestion and intention mining. Opinion mining. Existing works on opinion mining are performed at document, sentence and aspect level. At the document (Pang et al., 2002; Turney, 2002) and the sentence level (Wiebe et al., 1999), the task is to detect the polarity of a given document or sentence. At the aspect level, opinions are extracted as a tuple of entity, aspect, sentiment, …

Content truncated to 500 words