ABSTRACTWith and filtering out spam reviews [1].Opinion Mining and

ABSTRACTWith the explosive growth of online social media, people are able to express and share their opinions with each other on a common platform. People’s opinions and experience become the most important source of information in Sentiment Analysis process for decision making. The effectiveness of opinion mining relies on the availability of credible opinion for sentiment analysis. Unfortunately, spammers write fake reviews which can be positive or negative opinions in order to promote their services or damage the reputation of their competitors’ services. Spam reviews may mislead potential customers and affect their experience and influence their ideas so it is a must to detect and eliminate these fake reviews so as to prevent deceptive potential customers. In this paper, an enhanced approach for feature level Sentiment Analysis of social networks is presented. The proposed approach is developed to overcome limitations of previous work performed on feature based sentiment analysis process and assign accurate sentiment score for each feature in social reviews by considering Negation Handling and Spam Reviews Detection. The results indicate that our approach is accurate in classifying features and in specifying users’ opinions and attitudes.KEYWORDSOpinion mining, Sentiment analysis, Reviews, Feature, Negation handling, Spam reviews, Deceptive.1. INTRODUCTIONIn the past years, the World Wide Web (WWW) has come to be a massive source of user-generated content and opinionative data. Several websites mainly social networks such as Twitter and Facebook encourage users to publicly express their feelings in their daily interactions and exchange their views, suggestions and opinions related to products, services, etc. The increased popularity of these sites resulted in a huge collection of people opinions and valuable information on the web in much unstructured manner for the decision making process 1.The extraction of the useful content from these sites and the analysis of it became a challenging task. This clearly states that there is a need to work towards these challenges as it has opened up several opportunities for future research in the area of Opinion Mining and Sentiment Analysis for handling negations, hidden sentiments identification and filtering out spam reviews 1.Opinion Mining and Sentiment Analysis is an extension of data mining which utilizes natural language processing techniques to extract and classify people’s opinions expressed on entities or features of entities in an automatic fashion 2.An entity can be a product, service, person, organization or event while features are attributes or components of the entities. This helps in improving and addressing how many users support a sentiment whether it is in negation or in favor of a topic. It also helps in identifying the like-minded cluster of people and their relationship through sentiment mining 3.Sentiment analysis can be performed at different levels: document level where the whole document is classified as positive, negative or neutral, sentence level where the whole sentence is classified as2positive, negative or neutral, and aspect level where the whole document or sentence is classified as positive, negative or neutral for each feature present in the document or the sentence 9.Aspect-level sentiment analysis yields very fine-grained sentiment information which can be useful for applications in various domains due to the fact that when a user writes an opinion, it does not mean that the user likes or dislikes everything about the product or service. That is, users give points of view of several aspects that can be positive and negative. This information is important not only for users but also for companies because it supports a decision concerning buying a product, and it can serve as the basis to improve the products and services, respectively 16.There are several challenges that face the opinion mining and sentiment analysis process. These challenges become obstacles in detecting the accurate sentiment polarity 4.One of these main challenges is the detection of fake reviews. Spammers may write fake reviews which can be positive opinions to promote their products or negative opinions to damage the reputation of their competitors’ products. Spam opinions may prevent customers and corporations accomplishing real conclusions about the products. Therefore, it highly influences the e-commerce enterprise 7.Spam review detection is needed to identify and filter out spam opinions to provide real and trustful review services 5.Another challenge is the presence of negation words as they change the polarity of the text. Therefore negation words must be taken into consideration in sentiment analysis for correct polarity computation.The main contribution of this work is developing a feature based Sentiment Analysis approach to classify the reviews based on their features and assign accurate sentiment score for each feature in social reviews by considering negation handling and spam reviews detection.The rest of this paper is organized as follows: Section 2 presents an overview on related works. Section 3 describes the proposed approach. Section 4 presents the experimental study conducted and the results collected out of the proposed approach. Finally, Section 5 concludes the paper.