STUDY REPORT: RESTURANT REVIEWS SENTIMENT ANALYSIS SUMMARY Opinion based restaurant reviews are common used by consumers to decide which restaurant products they are going to use. The stages of sentimentally analyzing a restaurant are three main stages which are gathering of data, aspect classification which involve aspect identification and grouping and finally determining the focus of the aspects obtained. Even though each stage is faced by its own set of challenges we have seen the way the challenges experienced in the second stage (aspect classification) can be minimized by the use of machine learning techniques, lexical and natural language techniques which all do ease the process identifying aspects from mined reviews. INTRODUCTION …show more content…
Mining of opinions from customer reviews can be done at three levels which include the sentence level, document and the aspect level. In this paper we will focus on the aspect based mining of opinions since the sentence level focuses on classifying each sentence as either positive, neutral or negative and the document level classification aims at classifying the whole document review as negative or positive. Clearly restaurant reviews are a combination of opinions which would not be mined by the sentence and the document level analysis, only the aspect based would be able to mine the mixed opinions (Bing, 2012). Identifying aspects, identifying words and detecting orientation are the main tasks involved in aspect based mining of reviews. A good example of the aspect mining is a restaurant with a good environment but with bad food. The initial step would be to identify the aspects of the review which would be the food and the environment. Next would be to identify the words related to aspect which would be the opinion, in our case it would be bad and good. The final step would be detecting the focus the opinion by knowing the orientation expressed by the words to be either positive or negative. SentiWordNet helps a lot in valuing each opinion as positive or negative (Ohana, Bruno and Brendan,