Topic Drift in Comments Section of Articles : Natural Language Processing | IIT Kharagpur | Jan ‘16 - Apr ‘16 |
I tudied the phenomena of Topic Drifting in the comments section of an online article by identifying the comment which deviates from the topics touched by the article or discussion in the prior comments. I applied LDA (Latent Dirichlet Allocation) Model to generate the topics associated with the content of the article, Thus, tagging article sentences to the corresponding topic. I finally employed cosine similarity metric to map comments to the topics generated from the article to filter out the true negatives.