Carrot2

Carrot2 is a programming library for clustering text. It can automatically discover groups of related documents and label them with short key terms or phrases.

Carrot2 can turn, for example, search result titles and snippets into groups.

Web Search Clustering organizes search results from public search engines into clusters; offers treemap- and pie-chart visualizations of the clusters.

• Web: Web search results provided by etools.ch. Extensive use may require special arrangements with the owner of the etools.ch service.
• PubMed: abstracts of medical papers from the PubMed database provided by NCBI.
• Local file: content read from a local file in Carrot2 XML, JSON, CSV or Excel format.
• Solr: queries an Apache Solr instance.

Clustering Workbench clusters content from local files in JSON or Excel format, Solr or Elasticsearch; allows tuning of clustering parameters and exporting results as Excel or JSON.

• Lingo: creates well-described flat clusters. Does not scale beyond a few thousand search results. Available as part of the open source Carrot2 framework.
• STC: the classic search results clustering algorithm. Produces flat cluster with adequate description, very fast. Available as part of the open source Carrot2 framework
• k-means: base line clustering algorithm, produces bag-of-words style cluster descriptions. Available as part of the open source Carrot2 framework