Seniment Analysis on Hindi reviews.

Requirements:

1. python3 (Anaconda environment is preferred)
2. Scikit-learn
3. Numpy, Pandas
4. NLTK
5. googletrans
6. Pickle
7. Codecs

Problem:

We have used three approaches to classify the sentiment of Hindi reviews as positive or Negative.

1. Resource Based Semantic Analysis using HindiSentiWordnet.---> In this approach we used Hindi Sentiwordnet to classify the review's sentiment.
2. IN language Semantic Analysis. : This approach is based on training the classifiers on the same language as text.
3. Machine Translation Based Semantic Analysis. : In this approach we train the classifier on English reviews and for testing, we translate the Hindi reviews into English using Googletrans api and then we classify the Sentiment of review.

Dataset Used:

We have used a total of 1000 Hindi movie reviews for the Sentiment Analysis. We have taken 250 labeled reviews from the dataset of IIT- Bombay which contain 125 positive and 125 negative Hindi movie reviews. In addition, we have manually collected 750 reviews from a Hindi movie review website (Jagaran.com) and labeled them as positive or negative manually. Out of 750 reviews collected manually, 375 reviews are positive and the rest 375 are negative review. For Machine Translation based approach, we also need english reviews, so We have used NLTK dataset for english reviews.

Files Description:

classifiers.py : This module is used to do In-language classification. It applies different types of classifiers on the featureset generated using Bag of word model with feature value as TermFrequency or Term-Frequency-Inverse-Document_Frequency(TFIDF).

dbn_neuralnet.py: This module is used to do In-language classification of sentiment using Deep belief network(DBN) as a classifier.

MachineTranslationBasedApproach.py: This module is used to do Machine Translation Based Semantic analysis using Decision Tree as a classifier. We have used TF or TFIDF as a feature.

ResourceBasedSentimentClassification.py: This module is used to do Resource based sentiment classification of hindi reviews using HindiSentiWordnet as a resource.

UnigramTfFeatureGeneration.py: This module is used to generate Unigram+Tf Featureset of reviews. UnigramTfidfFeaturesetGeneration.py: This module is used to generate Unigram+Tfidf Featureset of reviews.

pos_hindi.txt: This contains positive hindi reviews of dataset. Reviews are seperated by $. neg_hindi.txt: This contains negative Hindi reviews of dataset. pos_english.txt: This contains positive english reviews. These are used in Machine Translation based approach. neg_english.txt: This contains negative english reviews.

dbn_outside: This is a directory Which contain deep belief network implementation.

How To RUN:

1)Run on terminal 'python ResourceBasedSentimentClassification.py' to do the sentiment classification through HindiSentiwordnet. It is called Resource Based Semantic analysis.

2)Run on terminal 'python classifiers.py' to do In language Semantic Analysis.

3. Run on terminal 'python dbn_neuralnet. to do In-language classification through Deep Belief Networks.

3)Run on terminal 'python MachineTranslationApproach.py' to do Machine Translation Based Semantic Analysis.