Design of a Sentiment Lexicon for the Greek Food and Beverage Sector

Abstract

Sentiment Analysis is a computational method aiming to extract opinions/evaluations of individuals for an entity such as a product, a service etc. In social media networks and other online sources, as for example the food websites, sentiment analysis is able to identify all possible terms, such as simple words, combinations of words, or phrases (pre-processing stage) that can be used to express the feelings of a user for a specific entity. Then, by considering the characteristics of these media, such as time sensitivity, text size limitation and unstructured expressions, converts them giving a positive or negative significance. For the analysis, a set of linguistic, statistical and machine learning techniques are usually considered to structure the information contained in text sources. The main purpose of this paper is twofold. First is to provide a literature review of the sentiment analysis techniques and second is to design a sentiment lexicon as a preliminary step to further analyze the customers’ reviews of some leading companies in the Greek Food and Beverage industry as they uploaded in the most common Opinion Social Networks in Greece (fb). Existing research has focused mainly on the recognition on English characters, while to our knowledge, limited research papers have been published so far concerning the Greek language, concentrating mainly on the banking and financial sector, neglecting contributions on food industry. Note that since significant portion of online text-based Greek communications ignore the rules of spelling and grammar the study takes into account this trend and improves the calculation of a sentiment score accordingly. As appears, the findings are expected to contribute in the design issues of a sentiment lexicon particularly devoted to the Greek food and beverage industry and to be used for further analysis.