Data Analysis & Sentiment Analysis for Unstructured Data

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476

Data Analysis & Sentiment Analysis for Unstructured Data Mukesh Yadav, PIIT NEW PANVEL Varunakshi Bhojane, PIIT NEW PANVEL ABSTRACT Data analysis means analyzing the information in order to draw the conclusion and understand the overall situation. In order to understand data machine needs to understand what are sentiments from the input like movie review, news review, product review, comments from blogs or posts or any other social website and give output as positive or negative review. Various algorithms and classifiers are present for sentiment analysis. This paper gives a survey of various classifiers and the approaches used. Keywords Data analysis, Qualitative data, Quantitative data, Sentiment, Sentiment Analysis, Natural Language Processing.

INTRODUCTION Data analysis means analyzing information in ways that reveal relationships, patterns, trends found within it. It tells us to what level we can trust the answers we are getting by comparing our information with others to get drawing the conclusion from the data. There are two kinds of data to work with. One is Qualitative data which refers to the information which is collected or can be translated into numbers, which can be displayed and analyzed mathematically. Another is Qualitative data which are collected as descriptions, anecdotes, opinions, quotes, interpretations, etc. Data analysis popular tool used is excel in which we can sort in ascending and descending order, filter our data that meet certain criteria, conditional formatting to highlight cells, display charts, extract the significance from a large and detailed data set using pivot tables, tables to analyze our data quickly and easily, Solver that uses techniques from the operations research to find optimal solutions for all kinds of decision problems, Analysis ToolPak which is an add-in program that provides data analysis tools for financial, statistical and engineering data analysis. Text information contains facts and opinions. Facts are entities, events and their properties also called as objective expressions. Opinions describe a person’s feelings, sentiments, appraisal’s towards entities, events and their properties. Sentiments are important and crucial whenever we need to make a decision we want to know other’s opinions. Whether the decision made or yet to be made is matching the opinions of the others in order to avoid self loss. One of the reasons for the lack of opinions is the fact that there was little opinionated text available before the World Wide Web. Before the web, when a person needed to make a decision, he or she asks for opinion from friends and family. But the web changed the way that people express their views and opinions by posting reviews of products at merchant sites, on Internet forums, discussion groups and blogs which are collectively called as the user-generated content. This online word of mouth behavior represents new and measurable sources of information with many practical applications. Sentiment analysis is the name given to the process of identifying the people’s attitudes and emotional states from language. In Natural Language Processing (NLP), sentiment analysis is an automated task where machine learning is used to rapidly determine the sentiment of large amount of text or speech. Applications include tasks like determining how excited someone is about an upcoming movie, political party with people’s likeliness to vote for that party, converting written restaurant review into 5-star scale across various categories like food

35

Mukesh Yadav, Varunakshi Bhojane

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476 quality, service and value of money. Automated sentiment analysis is the process of training a computer to identify sentiment within content through Natural Language Processing (NLP). Automated sentiment analysis will never be as accurate as human analysis, because it doesn’t account for subtleties of sarcasm or body language.[1]

LITERATURE SURVEY While most researchers focus on assigning sentiments to documents others focus on more specific tasks like finding the sentiments of words (Hatzivassiloglou & McKeown 1997), subjective expressions (Wilson et al. 2005; Kim & Hovy 2004), subjective sentences (Pang & Lee 2004) and topics (Yi et al. 2003; Nasukawa & Yi 2003; Hiroshi et al. 2004). These tasks analyse sentiment at a fine grained level and can be used to improve the effectiveness of a sentiment classification, as shown in Pang & Lee (2004). Instead of carrying out a sentiment classification or an opinion extraction, Choi et al. (2005) focus on extracting the sources of opinions, e.g., the persons or organizations who play a crucial role in influencing other individuals’ opinions. Various data sources have been used, ranging from product reviews, customer feedback, the Document Understanding Conference (DUC) corpus, the Multi Perspective Question Answering (MPQA) corpus and the Wall Street Journal (WSJ) corpus. To automate sentiment analysis, different approaches have been applied to predict the sentiments of words, e pressions or documents. hese are atural anguage rocessing and pattern-based i et al. asukawa i iroshi et al. onig Brill 2006), machine learning algorithms, such as Naive Bayes (NB), Maximum Entropy (ME), Support Vector Machine (SVM) (Joachims 1998), and unsupervised learning (Turney 2002). Table 1. Summary of Literature Survey Sr. No.

Title of the paper

Author & Year of publication

Observations/Remarks

1.

Thumbs up? Sentiment classification using machine learning techniques

Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan, 2002

2.

Thumbs up or thumbs down? Semantic orientation applied to unsupervised classification of reviews

Peter D. Turney, 2002

3.

Sentiment analysis: capturing favorability using natural language processing

Tetsuya Nasukawa and Jeonghee Yi,2003

4.

Determining the Sentiment of opinions

SOO Kim and Hovy, 2004

This paper used Naïve based classification, Maximum entropy and support vector machine. N-gram approach along with POS information is used to perform machine learning for determining the polarity. Tried different variations of n-gram, unigram, bigram, position and POS. SVM gave the highest accuracy with unigram feature. Accuracy 77-82.9 This paper used tag patterns with a window of maximum 3 words. In his experiment adjective(JJ),adverb(RB), single common noun(NN), plural common noun (NNS) were considered. Given a phrase, PMI (Point-wise Mutual Information) is calculated. Algorithm follows 3 steps: extract phrases containing adjective or adverbs, calculate semantic orientation & average semantic orientation. Accuracy of 65.8-84 Objective is to assign topic (subject term & paragraph) sentiments. It uses NLP and pattern based model. Data source taken were web pages & camera reviews. POS tagging is used to disambiguate some expressions. Syntactic parsing is used to identify relationships between sentiment expressions and subject term. Data set is present. Accuracy of 94.3 &94.5, Recall value 28.6 & 24. Describe an opinion as a quadruple [Topic, Holder, Claim, Sentiment] in which holder believes a Claim about the topic & in many cases associates with the sentiment. The system operates in 4 steps-select sentence that contain both topic & holder candidates, holder based regions of opinions are delimited, sentence sentiment classifier calculates polarity of all sentiment

36

Mukesh Yadav, Varunakshi Bhojane

E.

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476 bearing words, system combines them to produce holders sentiment for the whole sentence. It uses probabilistic based model. DUC 2001 corpus is used. Accuracy between 75.6-81 5.

A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts

Pang, Bo, and Lillian Lee, 2004

6.

Sentiment analysis: Adjectives and adverbs are better than adjectives alone

Farah Benamara, Carmine Cesarano, Antonio Picariello, Diego Reforgiato, and V. S. Subrahmanian, 2007

7.

Towards Enhanced Opinion Classification using NLP Techniques

Bakliwal, Akshat, et al., 2011

8.

Sentiment classification based supervised latent n-gram analysis

Bespalov, Dmitriy, et al., 2011

9.

A system for real time Twitter sentiment analysis of 2012 US Presidential election cycle

Hao Wang et al, 2012

10.

Large-Scale Sentiment Anlaysis for News and blogs

Namrata Godbole, Manjunath Srinivasaiah, Steven Skiena, 2007

on

Uses categorization technique only for the subjective portions of the document. Foe extracting the subjective portions an efficient technique is used for finding minimum cuts in graphs. It is a 2 step process: one is labeling the sentences in the document as either subjective or objective, discarding the objective sentences. Second, applying the standard machine learning classifier to the resulting extract of subjective sentences. AAC (Adverb-adjective combination) based SA technique is used. Minimizers are present which are small number of adverbs such as “hardly” that actually has a negative effect on sentiment. Unary AAC & Binary AAC scoring function is used to take input. There are 3 AAC scoring algorithms - Variable Scoring, Adjective Priority Scoring & Adverb First Scoring. Algorithm for scoring the strength of sentiment on a topic. It uses 2 approaches. First, Simple N-Gram matching. Second, POS tagging N-Gram matching. A scoring function which gives the priority to trigram matching followed by bigrams and unigrams is proposed. Normalization is done between 0 to 1 scale. It proposed an efficient embedding for modeling higher-order (ngram) phrases that projects the n-grams to low-dimensional latent semantic space, where a classification function can be defined. They utilize a deep neural network to build a unified discriminative framework to estimate parameters of the latent space & classification function. Framework is applied to largescale sentimental classification task. There are two large data sets for online product reviews. System architecture for real time processing data is given. In which real time data is taken, preprocessing i.e. tokenization, Match tweet to candidates, Sentiment model, Aggregate by candidate, Visualization and Online human annotation i.e. interface. A system is designed that assign scores indicating positive or negative opinion to each distinct entity in the text corpus. System consists of a sentiment classification phase, which associates expressed opinions with each relevant entity and a sentiment aggregation and scoring phase, which scores each entity relative to others in the same class.

PROBLEM STATEMENT It is normal for us to consult our dear ones whenever we plan to do something significant, for instance buying a home, going for higher education or choosing a profession. Getting opinions from other people aids us in decision making. Opinions are important when it comes to making a stable decision. Opinions about something makes the hold of the claim stronger thus allowing us to make a suitable and better decision. Sentiment analysis and opinion classification play an important role in predicting people’s views. The current trends in SA focus on assigning a polarity to subjective expressions in order to decide the objectivity/subjectivity orientation of a document or the positive/negative polarity of an opinion sentence within a document. Additional work has focused on the strength of an opinion expression where each clause within a sentence can have a neutral, low, medium or a high strength. Most of the work in the past on sentiment analysis deals with determining the strength of subjective expression within a sentence or a document using the parts of speech.

37

Mukesh Yadav, Varunakshi Bhojane

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476 Sentiment analysis tries to classify opinion sentences in a document on the basis of their polarity as positive or negative, which can be used in various ways and in many applications for example, marketing and contextual advertising, suggestion systems based on the user likes and ratings, recommendation systems etc. In the past machine learning techniques have been used to classify sentiments into positive and negative classes depending on their polarity. Benamara et al.[7] used adjective adverb combinations to assign positive and negative scores to sentiments. They generated three axioms to score adverbs of degree. Their methods give a Pearson correlation as high as 0.47. This technique gives a high precision and recall than previously developed algorithms. The axioms can be extended to other categories of adverbs, to study other syntactic constructions and to study the impact of guidelines. Bakliwal et al.[8] have used two NLP approaches like simple Ngram matching and POS tagged Ngram matching to assign polarity to opinions. They used their approach on product and movie reviews. They achieved a maximum accuracy of 76.3. Keeping in mind the above two approaches we can develop a new algorithm and achieve better accuracy. Our goal would be to minimize the human effort required to read a huge content and get a positive or negative opinion out of it. Using the best suitable approach and different classifiers to solve the problem of sentiment analysis.

CLASSIFIERS To implement machine learning algorithms on document data, we use the following standard bag-offeature framework. Let be a be a predefined set of features that can appear in a document. E amples include the word “still” or the bigram “really stinks”. et be the number of times occurs in document Then, each document is represented by the document vector. [2] =

(1)

1. Word Sentiment Classifier In this we assemble small amount of seed words by hand. Sorted by polarity into two lists i.e. positive and negative. Add words from WordNet. There are synonyms and antonyms. We assume synonyms of positive words are mostly positive and antonyms mostly negative. For example, the positive word “good” has synonym “virtuous, honorable, righteous” and antonyms “evil, disreputable, unrighteous”. Antonyms of negative words are added to the positive list, and synonyms to the negative one. We use synonym & antonym because it is much simpler than the corpus. But not all synonyms and antonyms could be used. This indicates the need to develop a measure of strength of sentiment polarity to determine how strong a word is positive and how strong it is negative. Given a word WordNet is used to obtain the synonym set of unseen word to determine how it interacts with sentimental seed lists. We compute c

(2)

where is a sentiment category (positive or negative), is the unseen word and are the WordNet synonyms of Problems is that is it difficult to pick one sentiment category without

38

Mukesh Yadav, Varunakshi Bhojane

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476 considering conte t. Unigram model is not sufficient. For e ample, “ erm limits really hit a democracy” says rof. Fenno, the multi meaning word “hit” was used to e press negative point of view about term limits. If such combinations occur adjacently, we can use bigrams or trigrams in the seed word list. But it is more difficult to identify the sentiment correctly, if one of the word falls outside the sentiment region. [5] 2. Sentence Sentiment Classifier It uses direct matching and opinion holder algorithm for identifying the topic. Near each holder we identify a region in which sentiments would be considered. Any sentiment outside such a region we classify as undetermined and is ignored. Sentiment region can be defines in various ways like Window 1 contain full sentence, window 2 contain words between Holder and Topic, window 3 contain window2 plus or minus 2 words, window 4 contain window2 to the end of sentence. Problem with this is that in a sentence there may be two different opinions and system determines which is the closest one. For e ample, “She thinks term limits will give women more opportunities in politics” expresses a positive opinion about term limits but the absence of adjective, verb and noun sentiment words prevents a classification.[5] 3. Naïve Bayes A Bayes classifier is a simple probabilistic classifier based on applying Bayes theorem with strong independence assumptions. A more descriptive term for the underlying probability model would be ”independent feature model”. In simple terms, a aive Bayes classifier assumes that the presence or absence of a particular feature of a class is not related to the presence or absence of any other feature. Depending on the precise nature of the probability model, NB classifiers can be trained very efficiently in a supervised learning setting. In many practical applications, parameter estimation for NB models uses the method of maximum likelihood; in other words, one can work with the NB model without believing in Bayesian probability or using any Bayesian methods. An advantage of the NB classifier is that it requires a small amount of training data to estimate the parameters necessary for classification. Because independent variables are assumed, only the variances of the variables for each class need to be determined and not the entire covariance matrix. One approach to text classification is to assign to a given document the class = . Naïve Bayes (NB) classifier is derived by first observing that by Bayes rule, (3) where P plays no role in selecting . To estimate the term , Naïve Bayes decomposes it by assuming the are conditionally independent given ’s class. raining method consist of relatively estimation of and , using add-one smoothing.[2] 4. Maximum Entropy Maximum entropy classification (MaxEnt, or ME, for short) is an alternate technique which has proven effective in a number of natural language processing applications (Berger et al., 1996). Nigam et al. (1999) show that it sometimes, but not always, outperforms Naive Bayes at standard text classification. Its estimate of takes the following exponential form:

39

Mukesh Yadav, Varunakshi Bhojane

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476

(4) where is a normalization function. defines as follows :

is a feature/class function for feature

and class ,

(5) For instance, a particular feature/class function might fire and only if the bigram “still hate” appears and the document’s sentiment is hypothesized to be negative. Importantly, unlike aive Bayes, MaxEnt makes no assumptions about the relationships between features, and so might potentially perform better when conditional independence assumptions are not met. [2] 5. Multi-Layer Perceptron (MLP) Multi-Layer perceptron (MLP) is a feed-forward neural network with one or more layers between input and output layer. Feed-forward means that data flows in one direction from input to output layer (forward). It can be implemented in Weka toolkit. [8] 6. Bagging Bagging is a bootstrap ensemble method which creates individuals for its ensemble by training each classifier on a random redistribution of the training set. Each classifier’s training set is generated by randomly drawing, with replacement, N examples where N is the size of the original training set; many of the original examples may be repeated in the resulting training set while others may be left out. Each individual classifier in the ensemble is generated with a different random sampling of the training set. This is a classifier that involves adaptive reweighting and combining to improve classification. This algorithm uses multiple sets for training. Each bootstrap set is used to train a different component classifier. The final classification decision is based on vote of each component classifier. The component classifiers are all of the same form. Certain classifiers become un- stable if there are small changes in training data, this also gives way to large changes in accuracy. Bagging improves recognition for unstable classifiers by smoothing over discontinuities. 7. Decision Tree Decision tree learning uses a decision tree as a predictive model which maps observations about an item to conclusions about the item’s target value. It is one of the predictive modeling approaches used in statistics, data mining and machine learning. More descriptive names for such tree models are classification trees or regression trees. In these tree structures, leaves represent class labels and branches represent conjunctions of features that lead to those class labels. In decision analysis, a decision tree can be used to visually and explicitly represent decisions and decision making. In data mining, a decision tree describes data but not decisions; rather the resulting classification tree can be an input for decision making.

40

Mukesh Yadav, Varunakshi Bhojane

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476 Decision tree learning is a method commonly used in data mining. The goal is to create a model that predicts the value of a target variable based on several input variables. An example is shown on the right. Each interior node corresponds to one of the input variables; there are edges to children for each of the possible values of that input variable. Each leaf represents a value of the target variable given the values of the input variables represented by the path from the root to the leaf. A decision tree is a simple representation for classifying examples. Decision tree learning is one of the most successful techniques for supervised classification learning. A tree can be ”learned” by splitting the source set into subsets based on an attribute value test. his process is repeated on each derived subset in a recursive manner called recursive partitioning. The recursion is completed when the subset at a node has all the same value of the target variable, or when splitting no longer adds value to the predictions. This process of top down induction of decision trees is an example of a greedy algorithm, and it is by far the most common strategy for learning decision trees from data. A decision tree is a classifier expressed as a recursive partition of the instance space. The decision tree consists of nodes that form a rooted tree, meaning it is a directed tree with a node called root that has no incoming edges. All other nodes have exactly one incoming edge. A node with outgoing edges is called an internal or test node. All other nodes are called leaves (also known as terminal or decision nodes). In a decision tree, each internal node splits the instance space into two or more subspaces according to a certain discrete function of the input attributes values. In the simplest and most frequent case, each test considers a single attribute, such that the instance space is partitioned according to the attributes value. In the case of numeric attributes, the condition refers to a range. Each leaf is assigned to one class representing the most appropriate target value. Alternatively, the leaf may hold a probability vector indicating the probability of the target attribute having a certain value. Instances are classified by navigating them from the root of the tree down to a leaf, according to the outcome of the tests along the path. In data mining, decision trees can be described also as the combination of mathematical and computational techniques to aid the description, categorisation and generalisation of a given set of data.[9] 8. SVM light Support vector machines (SVMs) have been shown to be highly effective at traditional text categorization, generally outperforming Naïve Bayes. They are large-margins, rather than probabilistic, classifiers, in contrast, to Naïve Bayes and Maximum Entropy. The basic idea behind the training procedure is to find a hyperplane, represented by vector that not only separates the document vectors in one class from those in the other, but for which the separation, or margins, is as large as possible. This search corresponds to a constrained optimization problem; letting (corresponding to positive and negative) be the correct class of document , the solution can be written as :=

41

Mukesh Yadav, Varunakshi Bhojane

,

0

(6)

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476 where the ’s are obtained by solving a dual optimization problem. hose vector such that is greater than zero are called , since they are the only document vectors contributing light to . So SVM package is used for training and testing, with all parameters set to their default values, after first length normalizing the document vectors. [2] SVM light is an implementation of Vapnik’s Support Vector Machine for the problem of pattern recognition, for the problem of regression, and for the problem of learning a ranking function.[12] The algorithm has scalable memory requirements and can handle problems with many thousands of support vectors efficiently. The software also provides methods for assessing the generalization performance efficiently. It includes two efficient estimation methods for both error rate and precision/recall. XiAlpha estimates can be computed at essentially no computational expense, but they are conservatively biased. Almost unbiased estimates provide leave-one-out testing. SVMlight exploits that the results of most leave-one-outs are predetermined and need not be computed. New in this version is an algorithm for learning ranking functions. The goal is to learn a function from preference examples, so that it orders a new set of objects as accurately as possible. Such ranking problems naturally occur in applications like search engines and recommender systems. Furthermore, this version includes an algorithm for training large-scale transductive SVMs. The algorithm proceeds by solving a sequence of optimization problems lower-bounding the solution using a form of local search. A similar transductive learner, which can be thought of as a transductive version of k-Nearest Neighbor is the Spectral Graph Transducer. SVMlight can also train SVMs with cost models. The code has been used on a large range of problems, including text classification, image recognition tasks, bioinformatics and medical applications. Many tasks have the property of sparse instance vectors. This implementation makes use of this property which leads to a Very compact and efficient representation. SVM light consists of a learning module and a classification module. The classification module can be used to apply the learned model to new examples. There are different options for choosing the kernel. Some kernels available are: linear (default), polynomial, radial basis function, sigmoid. 9. POS Tagger A Part-Of-Speech Tagger (POS Tagger) is a software that reads text in a language and assigns parts of speech to each word such as adjective, verb, noun etc. although generally computational applications use more fine-grained OS tags like ’noun-plural’. his software is a Java implementation of the log-linear part-of-speech taggers described in Toutanova et al.(2003) The tagger was originally written by Kristina Toutanova. Since that time, Dan Klein, Christopher Manning, William Morgan, Anna Rafferty, Michel Galley, and John Bauer have improved its speed, performance, usability, and support for other languages. There are several versions of POS tagger available for download. The basic download contains two trained tagger models for English. The full download contains three trained English tagger models, an Arabic tagger model, a Chinese tagger model, and a German tagger model. Both versions include the same source and other required files. The tagger can be retrained on any language, given POS-annotated training text for the language. The English taggers use the Penn Treebank tag set. The tagger is licensed under the GNU General Public License (v2 or later). Source is included. Source is included. The package includes components for command-line invocation, running as a server, and a Java API. The tagger code is dual licensed (in a similar manner to MySQL, etc.). Open source licensing is under the full GPL, which allows many free uses. [13]

42

Mukesh Yadav, Varunakshi Bhojane

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476

APPLICATIONS These are the applications of sentiment analysis. o In social media monitoring :  VOC to track customer reviews, survey responses, competitors, it is also practical for use in business analytics and situations in which text needs to be analyzed.  Computing customer satisfaction metrics : We can get an idea of how happy customers are with your products from the ratio of positive to negative tweets about them. o Identifying detractors and promoters  It can be used for customer service, by spotting dissatisfaction or problems with products.  It can also be used to find people who are happy with your products or services and their experiences can be used to promote your products. o In finance firms/markets  To forecast market movement based on news, blogs and social media sentiment.  To identify the clients with negative sentiment in social media or news and to increase the margin for transactions with them for default protection.  There are numerous news items, articles, blogs, and tweets about each public company. A sentiment analysis system can use these various sources to find articles that discuss the companies and aggregate the sentiment about them as a single score that can be used by an automated trading system. One such system is The Stock Sonar. This system (developed by Digital Trowel) shows graphically the daily positive and negative sentiment about each stock alongside the graph of the price of the stock. o Reviews of consumer products and services : There are many websites that provide automated summaries of reviews about products and about their specific aspects. A notable example of that is “Google roduct Search.” o Monitoring the reputation of a specific brand on Twitter and/or Facebook : One application that performs real-time analysis of tweets that contain a given term is tweetfeel. o Enables campaign managers to track how voters feel about different issues and how they relate to the speeches and actions of the candidates. o Applications in business domain Consider a question : “why aren’t customers buying our products?” or “why aren’t customers visiting our website?” We know the concrete data: price, specs, competition, etc. o In politics/political science; Evaluation of public/voters opinions. Views/discussions of policy. Law/policy making. Sociology; Psychology : investigations or experiments with data extracted from NL text.

CONCLUSION The experiments carried out on products/movie/news review dataset and obtained results by researchers showing the accuracy of the different approaches used. We conclude that parts of speech tagging gave the best result in classification of 76.6 percent as compared to all approach. The classifier that gave the best accuracy is the SVM light. The accuracy of classification varies according to different domains. Classifications can be done for more than two classes. It is possible to make a set of hybrid classifiers. Only one problem which can be faced while using the POS tagger is

43

Mukesh Yadav, Varunakshi Bhojane

International Journal of Engineering Technology, Management and Applied Sciences

www.ijetmas.com December 2014, Volume 2 Issue 7, ISSN 2349-4476 classifying neutral sentences, sentences that contain a negative as well as positive opinion which are difficult to classify and are misclassified many times. This issue can be taken up for classification. REFERENCES [1] Nitin Indurkhya, Fred J. Damerau, Sentiment analysis and subjectivity, In Handbook of natural language processing, Second Edition, pages 627- 630 [2] Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan. Thumbs up? Sentiment classification using machine learning techniques. In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7986, 2002. [3] Peter Turney. Thumbs up or thumbs down? Semantic orientation applied to unsupervised classification of reviews, pages 417424, 2002. [4] Tetsuya Nasukawa and Jeonghee Yi. Sentiment analysis: capturing favorability using natural language processing. In Proceedings of the 2nd international conference on Knowledge capture, K-CAP03, pages 70-77, 2003. [5] S.O Kim and E. Hovy, Determining the Sentiment of opinions, Coling04, 2004 [6] Pang, Bo, and Lillian Lee. A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts. Proceedings of the 42nd annual meeting on Association for Computational Linguistics. Association for Computational Linguistics, 2004. [7] Farah Benamara, Carmine Cesarano, Antonio Picariello, Diego Reforgiato, and V. S. Subrahmanian. Sentiment analysis: Adjectives and adverbs are better than adjectives alone. In Proceedings of the International Conference on Weblogs and Social Media (ICWSM), 2007. [8] Bakliwal, Akshat, et al. ” owards Enhanced Opinion Classification using echniques.” Sentiment Analysis where AI meets Psychology (SAAIP) (2011). [9] Bespalov, Dmitriy, et al. ”Sentiment classification based on supervised latent n-gram analysis.” Proceedings of the 20th ACM international conference on Information and knowledge management. ACM, 2011. [10] Hao Wang et al, A system for real time Twitter sentiment analysis of 2012 US Presidential election cycle. In Proceedings of the ACL 2012 System Demonstrations, pages 115-120, 2012 [11] Mostafa, Mohamed M. ”More than words: Social networks te t mining for consumer brand sentiments.” E pert Systems with Applications 1 . [12] T. Joachims, Making large-Scale SVM Learning Practical. Advances in Kernel MethodsSupport Vector Learning, B. Schlkopf and C. Burges and A. Smola (ed.), MIT-Press, 1999 [13] Kristina Toutanova, Dan Klein, Christopher Manning, and oram Singer. ”FeatureRich art-ofSpeech agging with a Cyclic Dependency etwork.” In roceedings of AAC , pp. 252-259

44

Mukesh Yadav, Varunakshi Bhojane