On-line help for Compendium Summarizer

COMPENDIUM is a Text Summarization system capable of generating generic informative summaries for English of a specific length provided by the user. As it can be seen in the user interface, COMPENDIUM relies on three different stages and techniques for generating a summary: - Redundancy Detection (RedD) is performed by employing textual entailment, - Topic Identification (TI) is achieved by word frequency, and - Relevance Detection (RelD) is obtained by the code quantity principle.

For generating a summary, once the source text has been inserted in the text box indicated with the name "Original", the techniques thorugh which the summary would be generated have to be selected. The available combinations of methods for generating the summary are:

  • TI
  • RelD
  • TI+RelD
  • RedD+TI
  • RedD+TI+RelD (this is the configuration which includes all options and it obtained the best experimental results)

The next step for summary generation is to decide its length, through the compression rate. This value can be selected from the list which contains predefined values ranging from 10% to 50%. In the case that with want another compression rate value, we have to select "other" and a text box will appear, where a new value can be indicated. Finally, the summary will appear in the "Summary" text box at the bottom, after pressing the button "Generate the summary".

More detail about COMPENDIUM system can be found in:

Lloret, Elena. 2011. Text Summarisation based on Human Language Technologies and its Applications. PhD. Thesis. Universidad de Alicante.

If you need to cite the system, please refer to the above reference.


This demo has been partially supported by the Spanish Government under the project (TIN-2009-13391-C04-01) and the Valencian Government (grant no. PROMETEO/2009/119 and ACOMP/2011/001). Many thanks also to Javi Fernández, José Manuel Gómez and Tatiana Vodolazova who have collaborated in the development of COMPENDIUM demo and its user interface.