Tuesday, January 24, 2006

Sift and separate.

link to original piece.

Reinforces the notion that what we need to know is probably known to us...only we do not know that we know.


Digging for data that can change our world

Research tools able to swiftly analyse masses of data could soon bring about advances that scientists up to now can only dream of

Julie Nightingale
Tuesday January 10, 2006

Scientific research is being added to at an alarming rate: the Human Genome Project alone is generating enough documentation to "sink battleships". So it's not surprising that academics seeking data to support a new hypothesis are getting swamped with information overload. As data banks build up worldwide, and access gets easier through technology, it has become easier to overlook vital facts and figures that could bring about groundbreaking discoveries.
The government's response has been to set up the National Centre for Text Mining, the world's first centre devoted to developing tools that can systematically analyse multiple research papers, abstracts and other documents, and then swiftly determine what they contain.

Text mining uses artificial intelligence techniques to look in texts for entities (a quality or characteristic, such as a date or job title) and concepts (the relationship between two genes, for example). In many ways, it's more precise and sophisticated than a search engine: it not only tracks down information against specified criteria but can also draw out relationships between hitherto unlinked bits of research (see below).

Initially, the centre is focusing on bioscience and biomedical texts to meet the increasing need for automated ways to interrogate, extract and manage textual information now flooding out of large-scale bio-projects.

"Biology is our primary focus as the government has identified that there's a big problem with all this information that nobody can handle," says Richard Barker, the centre's commercial manager.

The problems are compounded by the lack of standardisation. Scientists enjoy some licence in the nomenclature they use to register their discoveries. "In biology, for example, there are huge numbers of genes in the body and they are called different things by scientists, who are quite likely to use synonyms, abbreviations or acronyms," says Barker. "You can try to standardise but the problem is getting people to use them."

Text-mining tools in use include Cafetiere, an information extraction tool that annotates text with information about entities and the relationships between them. Termine, a tool for handling terminology, is being re-engineered by the centre so that it can deal with large volumes of data.

The centre, which is funded by science and higher education research councils and run by a consortium of Manchester, Salford and Liverpool universities, will act as a repository for such tools, as well as developing its own. One key task will be plugging the number of different tools for different tasks into one coherent framework.

"This infrastructure will allow many people's tools to work together in a mix and match way, the mix of which will depend on the intended application," says Barker.

In Manchester, the European Commission-funded BioMinT project is developing a text-mining tool to analyse documentation on DNA and protein sequence data spawned by the Human Genome Project, using the databases Swiss-Prot (devoted to protein sequences) and Prints (protein families) as sources. The tool, worked on by an interdisciplinary team of biologists and computer scientists, will handle queries of different kinds, retrieve relevant documents, extract the information and produce it either as a database entry or as a report.

"It will enable researchers to identify a range of information, such as the function of a particular gene or whether a protein is linked to a disease such as breast cancer, much more swiftly," says Terri Atwood, professor of bio-informatics at Manchester University, who is leading the project.

The collaboration across disciplines required in text mining is something academia is likely to see more of, she suggests. "One difficulty is bringing together the different communities of biologists and computer scientists. There's a huge language barrier between them in terms of the jargon they use.

"But this kind of interdisciplinary science will be a feature of many more scientific endeavours in the future, the more we use computers and want them to take over the things that humans do."

Backstory: How text-mining began

Text mining has been carried out since the mid-80s when the US academic, Prof Don Swanson, realised that, by combining information sliced from seemingly unrelated medical articles, it was possible to deduce new hypotheses. And this suggested that the existing body of medical literature might be full of previously unnoticed links.

Swanson took a manual approach: through intensive reading of the medical literature available, he extracted the proposition that fish oil might benefit patients suffering from the circulatory disorder, Raynaud's disease. The information had been implicit in the literature but no one had made the connection.

But now technology has speeded up the process, enabling large amounts of text - particularly abstracts and articles held in online medical databases such as Medline or open-access publishers such as BioMed Central - to be interrogated.

The more breathtaking results have included the discovery of new therapeutic uses for the drug Thalidomide to treat conditions such as chronic hepatitis C and acute pancreatitis and that chlorpromazine may reduce cardiac hypertrophy - enlargement of the heart leading to heart failure.


National Centre for Text Mining: www.nactem.ac.uk

BioMinT project: www.biomint.org


Post a Comment

<< Home