OpenEphyra

This page gives an overview of the OpenEphyra system, the project goals, and ways for you to participate. The documentation and a collection of tutorials are hosted on the Ephyra site. Downloads of new releases and a discussion forum can be found at SourceForge. If you would like to be notified of new releases and other important information regarding OpenEphyra, you can subscribe to our mailing list.

What is OpenEphyra?

OpenEphyra is the first open framework for question answering (QA). It retrieves accurate answers to natural language questions from the Web and other sources. OpenEphyra is hosted at SourceForge and published under the GNU General Public License (GPL). The framework comes with implementations of a number of algorithms that proved effective in CMU’s Ephyra system, which has participated in the TREC evaluation. It has been written entirely in Java and is thus platform-independent.

You can easily set up OpenEphyra on your own machine. Just download the latest version from SourceForge, unzip it, and run one of the scripts that are contained in the distribution. The only system requirements are a Java runtime environment (version 1.5 or later) and about 1 GB of free RAM. OpenEphyra is small, fast, and only requires a normal desktop (or laptop). Yet its performance is comparable to other state-of-the-art QA systems.

Why an open source QA framework?

  • QA systems become increasingly complex, and many researchers do not have the time or resources to build an end-to-end system. We invite you to use OpenEphyra as an easy-to-understand, yet powerful framework for your experiments.
  • Evaluation results reported for new QA approaches are often based on different underlying systems, and are therefore not comparable. By using a common framework, we can directly compare new techniques and show that one approach is preferable over another.
  • OpenEphyra can also be used for course projects and seminars. Students can download the framework and extend it with their own ideas to gain an insight into the field of QA.
  • And finally, we hope that other researchers will share their own components and help us to improve OpenEphyra and provide the QA community with a stronger baseline system for future experiments.

How is OpenEphyra different from CMU's Ephyra system?

The functionality of the two systems is very similar. We removed some third party tools, components that are OS dependent or have high hardware requirements, and code that is not well documented or tested. This was necessary to make OpenEphyra truly platform independent and useful to others, and to publish it under the GPL. However, our intention was not to cut any critical components, and in fact the performance of OpenEphyra comes very close to our evaluation system.

A note to researchers and developers:

We strongly encourage you to extend OpenEphyra and integrate new techniques and algorithms. OpenEphyra really is an open framework, intended to improve the collaboration between QA researchers. If you come up with improvements that may be useful to others, please share them with us so that we can include them in future releases. In this way you can also make sure that your code will be compatible with later versions of OpenEphyra. Contributions will be acknowledged on this website and in the readme file that comes with the download.

A note to teachers and lecturers:

We also encourage the use of OpenEphyra for course projects and seminars. Please contact us if you are interested in using the system for your course. We will do our best to answer your questions and assist you with any problems in a timely manner, although we cannot guarantee technical support.

Sponsors

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


We highly appreciate any feedback, comments, and suggestions for improvements. Please post to our forum or send and email to Nico Schlaefer.