Supervised Machine Learning Approach for Gender Disambiguation from User Generated Unstructured (Text) Documents

摘要

The rise of social media has led to an explosive growth in the size of data generated, data growth has undergone a renaissance. The pervasive espousal of social media into our daily lives has opened many opportunities for researchers to deep dive into human behavior. Many aspects of human behaviors have been explored using media data, example, detecting and monitoring mood state, forecasting sentiment analysis etc. Another important aspect of human behavior where a significant interest lies is identification of author identity. Predicting author characteristics, preferences and opinions helps answer many social science questions and support many commercial applications especially in e-commerce business. Identifying gender using author names and profile names by Twitter and Google are some examples of many advances in this area. Our work in this research takes it to the fore with ability to even classify anonymous users or authors. It is engrossed towards disambiguating author gender through lexical choice, choice of syntactic structure, capitalizing on linguistic nuances and textual meaning. The results of our research are quite promising and endure witness to the validity of approach.