###@Author Shivam Sharma(28shivamsharma@gmail.com) It is kaggle competition about mining sentiments of given test data. Link of the competition. My program is basic implementation of Naive Bayes.
How do I Run this program?
- First install R and packages listed below:-
- stringr
- openNLP
- NLP
- Run program in given sequence:-
First download the test & train data-set from kaggle link. And please let me know through email if data not found I will provide it by myself. And both test.csv & train.csv dataset in data folder of repositary. Then run following programs.
- Adjectives.R
- train.R
- Model.R
###Algorithm:- Algorithm contains 3 major steps:-
- Find features
- Find all probability required for calculating post-prior probability.
- Train on test data.
1.Find Features:-
I took adjectives(JJ/JJS/JJR) as features for my model. This approach is introduced by B. Pang[1] in
Empirical Methods in Natural Language Processing. I found all adjectives by combining all Phrase and
applying POS Tagging on the corpus. I found number of adjectives greater than 5000.
2.Building Naive Bayes:-
It is one best approach for text classification.It is probabilistic classification based on bayes
theorem with independent features.
(a) I calculated all classes occurring probability(P(class)).
(b) Then I calculated all features probability w.r.t each class.
3.Train:-
Applying above model to test data by selecting each sentence and calculating its sentiment.
###References:- [1] B. Pang, L. Lee, and S. Vaithyanathan. Thumbs up?Sentiment classification using machine learnin techniques. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 79-86, 2002.
[2] R. Feldman. Techniques and applications for sentiment analysis. Published in: Magazine Communications of the ACM Volume 56 Issue 4, April 2013 Pages 82-89.