2 past Research: Information Retrieval Models

My research interests lie in Machine Learning (ML) and its applications to Information Retrieval (IR), Text analysis and Data Mining (DM). These applied areas have historically been driven by empirical approaches. While these approaches have been quite successful in terms of performance, one of their major drawbacks is their lack of easy interpretability. It is my firm belief that machine learning approaches are a key towards solving some of the complex problems in text analysis and IR for the following reasons: (i) ML techniques are theoretically well grounded, making them easily interpretable, (ii) the theoretical framework of ML techniques acts as a guidance for future researchers to build extensions to existing models and (iii) the theoretical foundation also makes the models generalizable and portable across different tasks and domains. In line with this philosophy, my research has thus far been geared towards advancing the state-of-the-art technology in the above mentioned areas using theoretically well motivated, machine learning based approaches. While staying within the confines of this broad objective, I consciously made an effort not to limit myself to any particular school of thought and retained the flexibility to build models in diverse frameworks such as discriminative and generative, supervised and unsupervised approaches. I started out my research career as a graduate student in IR. I describe several theoretical contributions I made to this area in section 2. Not content with the knowledge I gained in my graduate study, I accepted a post doctoral fellowship in the Machine Learning department of Carnegie Mellon University (CMU), with a strong desire to sharpen my skills in Machine Learning. In section 3, I discuss my current research at CMU on topic modeling for data mining from large document collections. Section 4 describes my future research directions and section 5 concludes the statement with a discussion on my philosophy on other matters pertaining to a successful research career.