Home - GeeUnit/hw5-team07 GitHub Wiki

Welcome to the home of DEIIS Team #7!

The goal of this page is to describe the Q&A system project being development as part of the Fall 2013 Design and Engineering of Intelligent Information Systems (DEIIS) course at Carnegie Mellon University. The goal of the project is to improve automated question answering in the medical domain (specifically in Alzheimer's research).

The QA4MRE Challenge

The DEIIS project was based on attacking the Alzheimer's track of the QA4MRE challenge @CLEF 2013.

The challenge revolves around building a system that can correctly answer questions based on a passage about Alzheimer's disease. Successful attempts at the challenge incorporate elements of Machine Reading, Information Retrieval, and Natural Language Processing.

Please see the link for more details.

QA System

Our Q&A system code was forked from an early version of the system described in the Building an Optimal Question Answering System Automatically using Configuration Space Exploration (CSE) for QA4MRE 2013 Tasks (Patel et al 2013)

For more information, please see the Q&A Architecture page.

Analytics Pipeline

Our analytics pipeline is built using the Unstructured Information Management Architecture (UIMA) Framework. The pipeline employs: named entity extraction, part-of-speech tagging, sentence expansion, and synonym expansion among others.

For more details see the the Analytics Pipeline page.

Results

Our pipeline showed significant improvement over the baseline code. For CLEF 2012 Alzheimer test data, we improved roughly 100% with our analysis pipeline. Some of our tuning results for CLEF 2012 can be found in the results page.

Acknowledgements

We'd like to thank Professor Nyberg, and all the TA's who assisted in providing this course: Alkesh, Di, and Zi. We also wish the best of luck to the other teams!