EXTRACTING OPINIONS FROM REVIEWS - Anurag Kulkarni - Manisha Mishra -Raagini Venkatramani SCOPE OF THE PROJECT The project is primarily concerned with the development of an application which extracts opinions from reviews (available online) of various products viz. Identification of the opinion of the product. Determining the polarity of the opinion. Ranking the opinion based on their strength PROJECT SCHEDULING WORK TASK PLANNED START PLANNED COMPLETE ASSIGNED PERSON 1. Identify needs and benefits SEPTEMBER SEPTEMBER AK,MM,RV 2. Collection of reviews. OCTOBER OCTOBER RV 3. Creation of Vocabulary and dataset. OCTOBER OCTOBER RV,MM,AK 4. Design and Implementation. NOVEMBER NOVEMBER AK,MM,RV 5. Creation of visualizations. DECEMBER DECEMBER MM, AK Anurag Kulkarni: AK Manisha Mishra: MM Raagini Venkatramani: RV PROJECT IMPLEMENTATION Language Used : Java Platform Used : Eclipse Software for visualization : Tableau Logic Used : The project uses the idea of Naïve Bayes Classifier to classify the polarity of opinions. REVIEWS The project uses a collection of 40 reviews averagely for each product. The reviews are stored in a text file. They are gathered from sites www.amazon.com The project makes use of the reviews of the following two categories: Camera Hard-Disk Each item has five products in it. VOCABULARY A vocabulary is made manually on the basis of the reviews collected. Vocabulary contains the following : 1). Positive Words 2). Positive Phrases 3). Negative Words 4). Negative Phrases Example of Positive Words Vocabulary : 1). flawless 2). great 3). simple IMPLEMENTATION Algorithm Fetch review from the directory. Fragment the reviews in the sentences. Assign weight to the sentences based on the vocabulary. (Vocabulary contains positive words and phrases ,negative words and phrases) Sum up the weight of the sentences to get weight of the review. Sum up weight of the reviews to get overall product score. Rate of the product = number of reviews with positive score / number of reviews with negative scores + log(number of neutral reviews) OUTPUT OBSERVATION 1 OBSERVATION 2 OBSERVATION 3 CONCLUSION We have successfully analyzed reviews in the following manner: Identified opinions of various product. Determined the polarity of the opinion. Ranked opinions based on their strength. ACKNOWLEDGEMENTS We would like to thank Dr. Wengsheng Wu for guiding us in the project and making himself available all the times for clearing our doubts. Also we would like to thank our T.A Fei Xu for keeping himself available all the times to solve our doubts. BIBLIOGRAPHY AND REFERENCES BIBLIOGRAPHY: Introduction to Information Retrieval Christopher D. Manning, Prabhakar Raghavan, and Hinrich Schütze. Cambridge University Press, 2008. REFERENCES: Extracting Product Features and Opinions from Reviews , Ana-Maria Popescu, Oren Etzioni, Proceedings of HLT-EMNLP, 2005 http://reviews.ebay.com/POSITIVE-FEEDBACKuseful-WORDS-and-PHRASES-forBUYERS_W0QQugidZ10000000000733349 THANK YOU!!