My thesis research is in the field of computational sociolinguistics, in which I study manifestations of different types of social power (e.g., influence, hierarchy etc.) in the language and structure of interactions in order to build computational systems that can identify powerful and influential participants in them. Such systems have a variety of applications such as online marketing (e.g., targeting ads to influencers), market research (e.g., identifying opinion leaders), and intelligence and security (e.g., finding powerful people in suspicious online communities). This line of research can also help answer some of the fundamental questions in social sciences on how we humans interact, which will in turn help build socially-aware dialog systems for machine-human interactions.
As part of my thesis research, I have tackled and made significant contributions to various NLP/ML problems, such as introducing the formulation of mixed ngrams (NAACL2012) and the cascaded minority preference multi-class classification method (NAACL2013) both of which were shown to be useful in a number of other tasks later, reducing the error in dialog act tagging performance by 10% using these techniques, and proposing novel techniques using turn substantivity to detect shifts while modeling topics in interactions (EMNLP2014). Another area I actively research in is extracting extra-propositional aspects of meaning from text. I have built an automatic committed belief tagger (COLING2010), a modality tagger to extract different modalities (e.g., desire) expressed in text (ExProM2012), and systems for negation detection and uncertainty detection(CoNLL2010). I have also worked on relation extraction from biomedical domain as part of my Siemens (ECAI2012) and IBM Research internships, and on information retrieval as part of my Google internship.
- Curriculum Vitae (as of 07/2017)