TEXT ANALYTICS: Compelling products that pack a punch!

The process of examining and selecting a text analytics product often leads to a more strategic vision of what can be done, according to Tom Reamy, chief knowledge architect of KAPS Group. "Once an organization sees what text analytics can do for one business context, they often broaden their horizons," he says. "There is still something of a disconnect between text analytics and the more people-oriented part of KM that addresses issues such as how to create a knowledge sharing culture. Ideally they work together, because both have their strengths. A good example is the use of text analytics to semi-automate expertise location, which can help in team building."

Given the wealth of information and the pressing need to understand it, the outlook for text analytics is promising.  

Versatility in text analytics

"Text analytics is very much a platform technology," says Tom Reamy, chief knowledge architect of KAPS Group. "It can carry out a wide range of general and specific analyses."

Typically, text analytics applications rely on a combination of statistical analyses (such as word counts, clustering and pattern analysis) and linguistic analyses, which use an underlying categorization structure that is more context-sensitive and therefore more accurate. "The word ‘pipeline,' for example, has a different meaning in the oil industry and the pharmaceutical industry," adds Reamy. "If the software cannot distinguish between the two, the results will not be valuable."

The KAPS Group develops integrated semantic infrastructures for organizations, builds applications on top of the infrastructures and helps companies select the right text analytics software solution. "There is a bewildering variety of vendor offerings," Reamy says, "and it is important to determine which approach works best for an organization's content and use cases.  Some have large prebuilt vocabularies and look great out of box but are difficult to fine-tune, while others are very easy to use but are limited in their accuracy."

Once a software solution is selected based on the organization's requirements, the software is not necessarily plug and play. Multiple steps are often needed to prepare the text for analysis, such as marking the sections of documents, extracting concepts and building categories (which often benefits from human intervention). Depending on an organization's in-house resources, text analytics projects may be good candidates for hiring outside expertise.