Add Context to Content

In the past, context was king. The insights researchers were able to gain when conversing informally were extremely rich because each individual brought a powerful data processing source to the table: his or her own mind. Human brains are adept at making contextually relevant associations; they categorize information into loose groupings and can make intelligent connections that a structured database is incapable of. For example, a human would know immediately that the words "auto," "automobile," and "car" mean the same thing, or that a past experiment may be "kind of" like one being conducted in a current project.

But what happens when the available knowledge base includes an enormous breadth of sources, data formats, and locations? A researcher may be able to conclude that any files containing either the term aspirin or acetylsalicylic acid are relevant to his work, but would find it impossible to quickly or easily access the most important information if he had to actually read through thousands of pages of published literature, or manually search through hundreds of ELN documents to find it.

For example, one of the world's largest global pharmaceutical companies wanted to search a vast amount of unstructured content—ranging from external patents and journal articles to their own internal company documents—in order to identify and extract information related to specific new business opportunities for its existing intellectual property. The organization leveraged a services-based IT platform to integrate both the unstructured data sources, as well as an array of text-mining applications. The outcome of its work clearly indicated that while standard text mining applications were useful, scientifically aware text analysis methods were even more critical to success. These allowed researchers to scan the content for IUPAC, SMILES strings, and common/brand names of interest and quickly pinpoint the most contextually relevant sources of information. Standard text methods, such as FAST, cannot recognize chemical structures or biological sequences in the manner above and thus an integrated approach was the best solution. Without this ability to quickly integrate both an array of applications and content sources, the time and cost constraints involved in leveraging this valuable information would have been too high, and critical insights would have been missed.

The insights that lead to new breakthroughs are often hidden in a deluge of data, inaccessible to the researchers who need them, and disconnected from other relevant sources of information. In order to transform this data into the knowledge that drives discoveries, today's organizations need to bring back the contextually rich collaboration that existed at the company lunch table, but in a form more suited to the modern research environment. This requires a global, services-based IT architecture that supports cross-disciplinary data-sharing and integration, local information delivery, and scientifically aware search capabilities. As a result, organizations can take advantage of all relevant research sources like never before.

Frank Brown is chief science officer at Accelrys. He can be reached at FBrown@accelrys.com