The Ephyra Question Answering System

Ephyra is a modular and extensible framework for open domain question answering (QA). The system retrieves accurate answers to natural language questions from the Web and other sources. Ephyra has participated in the TREC evaluation and has been described in various scientific papers. The framework was developed by Nico Schlaefer at Universität Karlsruhe (TH) and later at Carnegie Mellon University, but many others have made valuable contributions over the past years. Special thanks go to (in no particular order):

• Guido Sautter, for his work on definitional QA and the NE recognizer.
• Jeongwoo Ko, who developed the Answer Generator (AG).
• Justin Betteridge, who developed the answer type classifier.
• Eric Nyberg, for his advice and for providing the environment for our QA research at CMU.
• Petra Gieselmann, Thomas Schaaf, and Alex Waibel, for their support in the early phase of the Ephyra project.

Ephyra also contains some components that have been adopted from the JAVELIN system.

We are currently working on OpenEphyra, an open framework for question answering that we would like to share with the QA community. Our goal is to give researchers the opportunity to develop new QA techniques without worrying about the end-to-end system. We believe that OpenEphyra can facilitate evaluations and comparisons of different approaches by providing a common platform for experiments. We also hope that OpenEphyra will be used for educational purposes, such as for computer science course projects. Please take a look at the OpenEphyra page for more information about this project, or visit our SourceForge project site to download the latest release.

Documentation

Here is a collection of useful resources that should help you to get started with Ephyra or OpenEphyra:

• Overview of the Architecture

• Task-Oriented Tutorials:
  • Adding Query Generation Techniques
  • Adding Unstructured Knowledge Sources
  • Adding Structured Knowledge Sources? (coming soon)
  • Adding Answer Extraction and Selection Techniques
  • Using Indri to Search a Document Collection
  • Running an Evaluation on TREC Data
  • Training the Answer Type Classifier
  • Adding Named Entity Taggers? (coming soon)
  • Using the Semantic Answer Extractor

Didn't find the information you need? Send an email to Nico Schlaefer. If you question is of general interest, we may add another tutorial.

• Outline of the Source Code (Ephyra)

• Javadoc Documentation (OpenEphyra)

• Publications:
  • Nico Schlaefer:
    A Semantic Approach to Question Answering.
    VDM Verlag Dr. Mueller, ISBN 3836450739, 2007.
    (amazon)
  • Nico Schlaefer, Jeongwoo Ko, Justin Betteridge, Guido Sautter, Manas Pathak, Eric Nyberg:
    Semantic Extensions of the Ephyra QA System for TREC 2007.
    In Proceedings of the Sixteenth Text REtrieval Conference (TREC), 2007.
    (pdf)
  • Nico Schlaefer, Petra Gieselmann, Guido Sautter:
    The Ephyra QA System at TREC 2006.
    In Proceedings of the Fifteenth Text REtrieval Conference (TREC), 2006.
    (pdf)
  • Nico Schlaefer, Petra Gieselmann, Thomas Schaaf, Alex Waibel:
    A Pattern Learning Approach to Question Answering within the Ephyra Framework.
    In Proceedings of the Ninth International Conference on TEXT, SPEECH and DIALOGUE (TSD), 2006.
    (pdf)

Developers

If you would like to contribute to the Ephyra project, please send an email to Nico Schlaefer to obtain write access to the Subversion repository and Trac project website. Once your account has been created, follow the Getting Started Guide for Eclipse Users to set up a working copy of Ephyra on your machine.

Sponsors

The Ephyra effort is supported in part by IBM Open Collaboration Agreement #W0652159.