Aspect based sentimental analysis of Sri Lankan Hotels

Abstract

Now a day’s people use to check review comments before they are going to experience new things most of the time. This is quite common specially in hotel areas. These reviews are the most convenient way for tourists to get an idea about the place they are going to visit and stay. Not only tourists, local people and service providers also refer these reviews. But the problem is people are so busy with day to day life and they don’t have much time to check the review manually. Specially people like hotel managers. So, people use to read top most reviews and get the idea. Most of the time that is not effective. Sometimes we get the idea that the place is good when we go through the first few reviews, but if we consider the overall comments it can be different. So, there should be an easy way to find this overall idea. For hotel management also can do their job easily if there is an automatic way to find about the status about their hotels. This research will try to provide a solution for the above-mentioned problem. Most of the text review provide by the customer are unstructured and not organized in a pre-defined manner. These texts are usually difficult, time-consuming and expensive to analyze, understand, and sort through. This is a complex analysis. These kinds of complex analysis can be done by using Sentiment analysis. It is also known as opinion mining. In sentiment analysis, unstructured information could be automatically transformed into structured data of public opinions. This is the most suitable one for mine the opinion of human. It is a type of text analysis that classifies the human generated text and makes decision by extracting and analyzing the text. It extracts the hidden knowledge from the unstructured texts exist in form of patterns and relationships. Opinion can be categorized as positive and negative, then measure the degree of positive or negative associated with the event. There are three type of opinion mining techniques. Document level, Sentence level and Feature /Aspect level. In document level it considers whole document or paragraph and provide the entire document or paragraph looks more positive than negative. In sentence level, it breaks the document into sentences and classify each of them as positive, negative or neutral. This is bit more specific here than document level. In Aspect level it is more specific than other two level. It extracts the specific feature and check the positivity or negativity according to that. 5 If we consider the review comments. User put those comments considering some specific features. As an example, if we consider hotel field, customer provide review considering food or rooms or other aspect. Customer or service provider who read these reviews also read considering these kinds of aspect. So, they need specific idea. If they want to know about food of that hotel. It is no point to provide the entire document or sentence level idea. Sometime entire sentence or document can be positive but idea about food can be negative. So, method use to analyze customer review should more specific. Because of that, in this research it used aspect level opinion mining.