Reframing Text Analytics with BI

I spent a pleasant and illuminating 90 minutes recently with Justin Langseth, president and co-founder of Clarabridge. Clarabridge sells text-mining software designed to integrate with business-intelligence tools. The company's solutions target both established text-analytics markets, such as life sciences, law enforcement and intelligence, as well as rapidly growing segments: marketing, CRM, reputation management and the like. But boosting Clarabridge is not my job, and, at least for those 90 minutes, it wasn't Justin's either. We did talk about the company's latest software release, but the bulk of our conversation, the helpful and illuminating part, was about the changing market landscape.Text mining, even if little known, has already made very significant contributions in those established categories, for instance by revolutionizing the pharmaceutical drug discovery process. Scientists can expeditiously mine biomedical literature for protein interactions to discern possible therapies, reserving time-consuming and expensive laboratory work and clinical trials for the few, most promising leads. Breck Baldwin of Alias-I goes so far as to state that "the Nobel Prize in Medicine will [soon] go to a researcher primarily focused on text analytics because of the discoveries made possible by the technology." (Breck related in e-mail that he said this in a December talk at the National Institutes of Health and did not get booted out.)

Yet this type of investigative analysis, per my conversation with Justin, is being pushed from the text-analytics center stage by a different variety of application. Instead of searching for the terrorist fish swimming in a sea of people, for the uncommon successful therapy, or for unusual and fraudulent financial transactions, the more recent of text-analytics adopters are most interested in statistically described characteristics of larger populations. They're looking for trends, for big picture rules, rather than for the few exceptional cases. They want a correspondingly different style of analysis, one that has served them well in working with conventional, numerical data sources. They want familiar BI interfaces and techniques, also adaptation of concepts such as ETL to their work with textual sources. And they want the software to deal with linguistic complexity and deliver information in the type of data structures they're already using for numerical data.

Clarabridge provides end-user tools for extracting information from text into databases for BI-integrated analysis. Attensity and ClearForest, among others, offer similar tools, also an option available from SPSS.

While mainstream analytics vendors have yet to deliver "BI for the Masses," for the vast majority of users, BI remains far more accessible than data mining and statistical techniques, which are often dismissed as esoteric. This simple fact suggests that BI-integrated text analytics will soon overtake offerings delivered through data-mining workbenches in popularity. This reframing of text analytics isn't a good or a bad thing. It's simply a welcome indicator of readiness for new challenges based on success meeting the old ones.

P.S. If you don't know about the Text Analytics Summits, which I chair, check them out now. The summits are slated for Amsterdam April 26-27 and Boston June 12-13. Note that I do not have a financial stake in registrations or sponsorships. I simply think they'll be great programs and want to encourage the widest possible participation.I spent a pleasant and illuminating 90 minutes recently with Justin Langseth, president and co-founder of Clarabridge. Clarabridge sells text-mining software designed to integrate with business-intelligence tools. But boosting Clarabridge is not my job, and, at least for those 90 minutes, it wasn't Justin's either. The bulk of our conversation was about the changing text mining market.