Opinion Mining and Sentiment Polarity on Twitter and Correlation between Events and Sentiment
Authors:
- Peiman BarnaghiPBPeiman Barnaghi
- Parsa GhaffariPGParsa Ghaffari
- John G. BreslinJohn G. Breslin
- 1 March 2016
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 52-57
Abstract
Twitter, as a social media is a very popular way of expressing opinions and interacting with other people in the online world. When taken in aggregation tweets can provide a reflection of public sentiment towards events. In this paper, we provide a positive or negative sentiment on Twitter posts using a well-known machine learning method for text categorization. In addition, we use manually labeled (positive/negative) tweets to build a trained method to accomplish a task. The task is looking for a correlation between twitter sentiment and events that have occurred. The trained model is based on the Bayesian Logistic Regression (BLR) classification method. We used external lexicons to detect subjective or objective tweets, added Unigram and Bigram features and used TF-IDF (Term Frequency-Inverse Document Frequency) to filter out the features. Using the FIFA World Cup 2014 as our case study, we used Twitter Streaming API and some of the official world cup hashtags to mine, filter and process tweets, in order to analyze the reflection of public sentiment towards unexpected events. The same approach, can be used as a basis for predicting future events.Keywords
This publication has 12 references indexed in Scilit:
- Automatic Unsupervised Polarity Detection on a Twitter Data Stream2014
- Techniques and applications for sentiment analysisCommunications of the ACM, 2013
- Sentiment Knowledge Discovery in Twitter Streaming Data2010
- The WEKA data mining softwareACM SIGKDD Explorations Newsletter, 2009
- Opinion Mining and Sentiment AnalysisFoundations and Trends® in Information Retrieval, 2008
- Large-Scale Bayesian Logistic Regression for Text CategorizationTechnometrics, 2007
- Thumbs up?2002
- BoosTexter: A Boosting-based System for Text CategorizationMachine Learning, 2000
- Maximizing text-mining performanceIEEE Intelligent Systems and their Applications, 1999
- Joining statistics with NLP for text categorization1992