How to implement Negation Features in SVM classification (NLP) using imdb Movie_Reviews corpus

Question

I am trying to understand Negation feature in NLP , so I thought to implement it. I am working on imdb movie review dataset. Consider I am having data as follows-

Movie was great but it's overly sentimental and at times terribly mushy , not to mention very manipulative but great action

From the above I can extract it's overly sentimental and at times terribly mushy as negative statement and now I am left with these choices-

• I extract the particular line till it ends with some punctuation and I simply remove this line from the positive statement and run SVM classifier in the rest of the content.
• I extract the particular line and label the line with negative and add it to the list of negative statements to train the same.

I am not sure that I am doing anything right here , so please suggest exactly how should I deal with negation features to improvise the classification.

I am working with scikit-learn svm.SVC() classifier

Answer 1

You can check this NRC Sentiment Analysis system for text classification using negation. It's very well explained. Also they claim their SemEval 2014 submission has major improvements on negation handling (I still haven't read it).

I assume you're solving a similar task on movie reviews so this must be what you're looking for.