Twitter Sentiment Analysis
Mert Kahyaoğlu
Instructor: Assoc. Prof. Dr. Bekir Taner Dinçer
NLP Course Project
AIM OF THE PROJECT
The aim of the project is to determine how people are feeling when they share something on twitter.
For example;
“ Final projeleri çok zevkli! ” => Positive
" Finaller çok kötü geçiyor :( ” => Negative
ROAD MAP
1.Collect Tweets
- Tweets;
- from Twitter API
- using Tweepy Helper Library
- Requirements
- Twitter Developer Application for authantication
2.Clean Tweets
- Hashtags ( # )
- Remove;
- Mentions ( @ )
- Links ( http://.. )
- Retweets ( RT .. )
- Limit ( length(word) > 2 )
Have you heard Galata by @cmylmz on #SoundCloud? https://soundcloud.com/cmylmz/galata
have you heard galata ?
- Lowercase
3. LABEL & ForMAT
- Emoticons ( 😃 )
- Label by;
- Adjectives
- Textual Emoticons ( ":D" )
- Format for Weka (.arff);
4.EXTRACT FEATURES
1. Apply Weka -> StringToWordVector Algorithm
"finaller çok kötü geçiyor" --> ["finaller", "çok", "kötü", "geçiyor"]
2. Attribute Selection
Selects attributes considering the individual predictive ability of each feature along with the degree of redundancy between them.
- Apply Weka -> CfsSubsetEval Algorithm
- Save new set of attributes as .arff file
Reduced terms from 930 to
5.CLASSIFY
- Boost Naive Bayes with AdaBoostM1 Algorithm
- What is Boosting?
- Train models (Naive Bayes) in a sequence
- Weights models according to performance
- New models are influenced by performance of previously built ones
- Encourage new model to focus on those cases which were incorrectly classified in the last round
- Combine the classifiers by letting them vote on the final prediction
USAGE
Results
Web APPLICATION
- Shows positive and negative tweets of your Twitter accounts