13 Jan 2003 - PharmaDM partnering BioMinT, the EC project for biological textmining.

The EC project "BioMinT" (QLRT-2001-02770, start date 1 January 2003) aims at the development of a generic text mining tool for the biological domain. More specifically, the goal of BioMinT is to:

1. semi-automate and therefore speed up the construction and update of SWISS-PROT , PRINTS and other protein databases while preserving their high standards of quality;
2. automate report generation from scientific abstracts and papers using the advanced information retrieval and -extraction features of the BioMinT text miner;
3. develop a semantic, concept-based information retrieval and extraction system for the biomedical literature.

Intended users are researchers at biotech & pharmaceutical companies, bench biologists and all bio-database curators.

BioMinT is a multidisciplinary project involving six partners: (key contribution between parentheses)

• SIB (Swissprot database)
• University of Manchester (Prints database and Precis, a tool producing protein reports based on Swissprot database annotations)
• University of Geneva (Info Retrieval/Info Extraction)
• University of Vienna (Info Retrieval/Info Extraction)
• University of Antwerp (Natural language processing)
• PharmaDM (Relational mining engines, AlphaDMax^TM)

An end user club has been established that involves some of the major players in the biotechnology and pharmaceutical industry. This end user club will allow researchers to give early feedback on issues such as (generic) requirements, ontology development, the BioMint GUI, etc..

The involvement of PharmaDM follows naturally from the description of the tasks ahead. In order to bridge the gap between word and concept, between syntax and semantics, some form of natural language understanding (NLU) technology in combination with domain-specific background knowledge is required. This will involve matters such as the customisation of natural language understanding techniques to biological texts and the incorporation of domain-specific knowledge into the information retrieval and extraction processes; matters that belong to PharmaDM's core expertise. Moreover, extraction of biological relations from text will certainly be better served by relational rather then propositional learning techniques. This is where PharmaDM's relational data mining technology will be of particular importance to the project.