Elif Aktolga Department of Computer Science 140 Governors Drive University of Massachusetts Amherst, MA 01003 Room 366 E-Mail: firstname at cs dot umass dott edu

I am a second year PhD student in the CIIR lab at the CS department of the University of Massachusetts at Amherst.

I am interested in fine-grained search issues that are typically dealt within the fields of Information Retrieval,
Question Answering, and NLP. My supervisor is Prof. James Allan.

Currently, I'm working on improving passage retrieval for question answering: (1) developing techniques for introducing more relevant passages for a given a question; (2) eliminating nonrelevant passages.

Some other projects I worked on in the past were

• Phrase detection from opinionated documents
• Clickthrough analysis of medical query logs
• Cross-lingual coreference retrieval by means of entity language models

This is my old homepage at the University of Osnabrück.

Here is an IR glossary which contains expressions and concepts from IR and related fields. The glossary is
incomplete and has to be improved, so I would appreciate feedback.

For those wondering which languages I speak and understand: German and Turkish (mother tongues), English, Hindi, and a bit of French.

I came to grad school because I am very interested in bringing computers closer to `understanding' natural language such that they can assist us with tasks like summarization, information retrieval, question answering, knowledge extraction, and machine translation. This task is so challenging because computers lack (1) a knowledge base: human beings acquire knowledge over years through education and experience. Can we build such a knowledge base that comprises `all the facts' and constantly extends itself by adding new information (learning)? (2) the ability to infer new facts; (3) the ability to interpret and reproduce text correctly. Among these issues I am most interested in parts 2 and 3. I believe that for success in these tasks techniques from the fields of information retrieval, natural language processing, and machine learning have to be synthesized.

I am very interested in exploring such techniques for Complex Question Answering, Knowledge Extraction and Passage Retrieval in particular. Eventually, I want to apply my research to building better chatterbots that can intelligently converse with people about advanced topics (and not mere `chatter').

1. Elif Aktolga. Pattern Matching Strategies for Peephole Optimisation. MRes Thesis, University of Sussex, UK, 2005. PDF, Slides


2. Elif Aktolga. A Java Planner for Blocksworld Problems. BSc Thesis, University of Osnabrueck, Germany, 2004. PDF

Back to top