Research-ESA is an open source implementation of Explicit Semantic Analysis. It supports the cross-lingual extension of ESA and allows the variation of the ESA model and parameters.

The Research-ESA library is available on the Research-ESA project site. It requires a local instance of the Wikipedia database (or alternative knowledge sources that can be used for ESA). It supports the following features:

• Indexing of concept spaces (using textual descriptions)
  • Support for Wikipedia in MediaWiki database format
  • Support for Wikipedia using Wikipediaminer
• Efficient implementation of ESA
  • Supports variation of association strength between text and concepts
  • Supports different pruning strategies for concept vectors
• Application of ESA for Information Retrieval
  • Implementation of inverted concept index
  • Implementation of different concept retrieval models

In addition to the Research-ESA library, the Research-ESA Web service allows to apply ESA to text using Wikipedia indexes from September 2009. On the client side, no local copy of Wikipedia is therefore needed.

Link to Research-ESA Web service configurator

Feature list of the Research-ESA Web service:

• Selection of concept spaces based on Wikipedia articles or Wikipedia categories
• Support of English, German, French and Spanish text
• Supports combination of concept indexes based on:
  • titles
  • redirect titles
  • article content
  • anchor text
• Selection of different ESA vector pruning strategies