Computational analysis of thematic blog data for sociological inference mining

This paper describes our proposed approach for computational analysis of thematic blog data through a novel combine of sophisticated Information Retrieval and Language Processing Techniques. We have implemented algorithms for Topic Modeling, Entity Extraction and Sentiment Classification with a view to draw sociologically relevant inferences from freeform unstructured social media data. Our experimental data comprised of more than 600 blog posts on the broader theme of `Discrimination, Abuse and Sexual Crime against Women´ collected during two discrete time periods. We have tried to extract some important inferences from the data such as key persons and organizations mentioned in the data, key themes encountered in the entire data collection, sentiment orientation inherent in the texts and variation in topic trends during the two discrete time periods. The results obtained are very interesting and validate the usefulness of our approach for computational analysis of social media data.