University centre achieves highest accuracy in text mining challenge

26 Oct 2013

The National Centre for Text Mining (NaCTeM) has won two first place rankings at the Biocreative IV challenge for developing software which can identify important information in biological text, accurately and automatically.

Text mining software facilitates the discovery, extraction and structuring of relevant knowledge from unstructured text. The output of text mining systems can enrich documents with semantic information, which can in turn be used to develop search systems that allow users to locate information of interest more quickly and efficiently than is possible using traditional search methods.

The National Centre for Text Mining (NaCTeM) in the School of Computer Science works on the development of text mining systems operating on various different subject areas. The quality of NaCTeM's text mining tools was recently demonstrated in BioCreative IV, which took place from 7 to 9 October 2013 in Bethesda, Maryland, USA. BioCreative IV is the latest in a series of challenges in which research groups from various different universities apply their text mining tools to extract a range of types of information from biology texts.

NaCTeM's text mining tools were ranked highest in two separate tasks involving the recognition of different types of concepts mentioned in texts. NaCTeM was ranked first out of 12 groups in the recognition of both chemicals and genes, and first out of 23 teams in another task involving the recognition of chemical names.