2016 IEEE Eighth International Conference on Advanced Computing (ICoAC) 978-1-5090-5888-4/16/$31.00@2016 IEEE 72 A SURVEY ON SENTIMENT ANALYSIS METHODS AND APPROACH Ms.A.M.Abirami Dept. of Information Technology Thiagarajar College of Engineering Madurai, Tamilnadu, India abiramiam@tce.edu Ms.V.Gayathri Dept. of Information Technology Thiagarajar College of Engineering Madurai, Tamilnadu, India gayathrivairam2015@gmail.com Abstract— Data Analytics is widely used in many industries and organization to make a better Business decision. By applying analytics to the structured and unstructured data the enterprises brings a great change in their way of planning and decision making. Sentiment analysis (or) opinion mining plays a significant role in our daily decision making process. These decisions may range from purchasing a product such as mobile phone to reviewing the movie to making investments---all the decisions will have a huge impact on the daily life. Sentiment Analysis is dealing with various issues such as Polarity Shift, accuracy related issues, Binary Classification problem and Data sparsity problem. However various methods were introduced for performing sentiment analysis, still that are not efficient in extracting the sentiment features from the given content of text. Naïve Bayes, Support Vector Machine, Maximum Entropy are the machine learning algorithms used for sentiment analysis which has only a limited sentiment classification category ranging between positive and negative . Especially supervised and unsupervised algorithms have only limited accuracy in handling polarity shift and binary classification problem. Even though the advancement in sentiment Analysis technique there are various issues still to be noticed and make the analysis not accurately and efficiently. So this paper presents the survey on various sentiment Analysis methodologies and approaches in detailed. This will be helpful to earn clear knowledge about sentiment analysis methodologies . At last the comparison is made between various paper’s approach and issues addressed along with the metrics used. Keywords—Data Analytics, sentiment Analysis, Decision making. I. INTRODUCTION DATA Analytics is an art of processing raw data to extract some reasonable information. Data Analytics is widely used in many industries and organization to make a better Business decision. By applying analytics to the structured and unstructured data the enterprises brings a great change in their way of planning and decision making. Data analysis is the process of verifying, cleaning, and transforming in order to retrieve useful information from the data. This information will be more helpful in suggesting business conclusions and decisions- making. Data Analysis has a variety of angles and methods that combines many techniques in order to provide better accuracy. One of the most popular methods of data analysis technique is data mining that mainly concentrates on modeling and discovery of knowledge for prediction process rather than descriptive purposes. predictive analytics is mainly used for predicting forecasting/classification where as text analytics make use of statistical, linguistic and structural techniques in order to retrieve information from text sources. This text sources are mostly in the form of unstructured data. Sentiment analysis (or) opinion mining plays a significant role in our daily decision making process. These decisions may range from purchasing a product such as mobile phone to reviewing the movie to making investments---all the decisions will have a great impact on the daily life. In ancient days before buying a product / service people will seek opinion from their friends, neighbors, etc. But in internet era it is easy to seek opinion from different people around the world. Now- a-days people before buying any product/service will make a glance on review sites (e.g. CNET), e-Commerce sites (e.g. Amazon, eBay) and social media (e.g. twitter) to get a feedback about the specific product (or) service in market. Sentiment Analysis makes use of 3 terms in order to fetch the sentiment .That is object and feature, opinion holder, opinion and orientation. Sentiment Analysis deals with several technical challenges such as object identification, opinion orientation classification, and feature extraction. Usually sentiment analysis can be performed using supervised and unsupervised learning such as naïve Bayes, Neural Networks, Support Vector Machine. Among these three techniques SVM is considered to be more suitable for sentiment Analysis. Sentiment classification can be performed in 3 stages such as • Document level • Sentence level • Feature level In document and sentence level the sentiment analysis make use of only a single object and extracts only a single opinion from the single opinion holder. But this type of assumptions are not suitable for many situations. Extracting sentiment for entire document/blog will not be efficient as extracting sentiment by considering aspects of each subject in the particular sentence. Fig 1.1: Sentiment Analysis concepts