ANALYSING SENTIMENTS OF ONLINE REVIEWS ON RESTAURANTS IN MALAYSIA USING PREDICTIVE TEXT ANALYTICS

This study aims to analyse the overall sentiments of online reviews on restaurants in Malaysia using predictive text analytics. As we know in opinion mining, sentiment analysis is a prominent technique in predictive text mining. It is a technique that categorises opinions in unstructured text format into binary classification (ie. good or bad). The authors attempt to go beyond the binary classification by viewing texts as empirical entities derived using the Term Frequency - Inverse Document Frequency (TF-IDF) weighting algorithm. These empirical entities, based on online reviews of restaurants in Malaysia, are then manifested into hypothetically defined constructs closely reflecting their thematic and semantic nature. The were 4914 customer reviews from restaurants across 20 towns and cities in Malaysia scraped off TripAdvisor.com using web crawler tools. Then a series of analytical tests were carried out. First the online reviews were parsed, filtered and clustered using SAS Text Miner. Then the online reviews underwent the TF-IDF process to identify significant terms and weightages were assigned according to their importance. The TF-IDF process resulted in a series of important nouns and adjectives from the text corpus. Using these weightages of nouns and adjectives, the authors went on to thematise these terms based on their semantic nature to manifest hypothetical constructs. These constructs were based on the Mehrabian–Russell Stimulus Response Model. Subsequently the authors tested the associations among the constructs using variance-based and covariance-based Structural Equation Modelling (SEM). The authors were encouraged by this exploratory methodological approach in formulating predictive text analytics using SEM. Results indicated that sentiments were generally positive towards restaurants and the important terms derived were price, hospitality, location, waiting time, availability of parking and size of food portion.