A Corpus Driven, Aspect-based Sentiment Analysis To Evaluate In Almost Real-time, A Large Volume of Online Food & Beverage Reviews

Nowadays, more than ever, customers have access to other consumers’ digital evaluations concerning the products or services that they have consumed. The use of online review websites, by the potential digital consumers, makes them aware of the choices they have. This, enables them to make comparisons between all the available products or services. However, the big volume of the opinionative data that is produced continuously, creates difficulties when being analyzed by stakeholders, mostly due to human’s physical or mental restrictions. In this research, web scraping combined with an aspect-level sentiment analysis using the corpus-based technique, approached methodologically the problem, by identifying not only the relevant information, but also the particular expressions and phrases that the reviewers use over the Internet. The purpose is to recommend a corpus-based, sentiment analysis web system for detecting and quantifying customers’ opinions which are written in Greek language and referred to the Food and Beverage (F&B) sector in almost real-time. The system consists of two modules that constructed using the aforementioned methods. As far as the web scraping module is concerned, the BeautifulSoup and the Requests libraries of Python programming language were used. For the constructing purposes of the corpus-based sentiment analysis module, 80,500 customers’ reviews are extracted (data set) from 6,795 companies which selected randomly from the most popular Greek e-ordering platform. The evaluated functions are the quality of food, the customer service and the image of the company. The extracted sentiment orientation terms and phrases from the customers’ reviews are used to form the corresponding dictionaries of the functions and the appropriate pattern of tags, in order to proceed in the sentiment classification. Finally, the system is tested in the dataset and the findings will be practical and significant, as not enough attention has been paid in sentiment analysis techniques used in combination with a nonEnglish, like the modern Greek language.