I need to perform sentiment analysis on topic modelling. I have used LDA with grid search search to identify the topic. The next step would be to perform sentiment analysis on it. How should I go about? What should be the training data and what should be the labels?


You should have a labeled training data from the outset for sentiment analysis.

For instance, each review should be labeled as 0 (negative) or 1 (positive).

If you do not have a labeled dataset, you cannot properly "train" the sentiment based on your topics.

Hence, with LDA, you can get features (i.e., topics) but you need targets (i.e., sentiments) already labeled as well.

If you do not have a labled dataset, you should artificially create it or use already-labeled toy datasets.

Below are examples of toy datasets

IMDb large movie review dataset

Twitter Sentiment Corpus


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.