Replacing a piece of technology that it formerly OEM’ed from elsewhere, SAS announced it has bought Teragram, a provider of text processing tools. The 40-person, Cambridge, MA-based company, provides natural language analytic tools that include the text categorization tools that SAS has lacked. Under the deal, it will be run as an autonomous business of SAS.

The Teragam deal brings together two companies that have not dealt heavily with each other in the past. In part that’s because SAS used to OEM a natural language processing tool from Inxight, a provider that was subsequently swallowed up by Business Objects. And in part, it’s because Teragram has tended to focus on media industry clients, such as NYTimes.com, that have not been traditionally SAS’s sweet spot.

For instance, Teragram’s tools provide the context-based searches that you find on the New York Times website. It adds context-based search to search engines such as FAST, which was just bought by Microsoft. And it provides the kind of mobile-based question and answer tools where you enter the query, it parses it, applies context, and based on its metadata map of content, spits out an answer. Those capabilities complement what SAS’s Text miner tools do, which is primarily pattern matching, but without context. So Teragram is a logical addition to SAS.

However, the lack of joint field experience or the fact that Teragram’s markets aren’t a 100% match for SAS aren’t show-stoppers for the deal. For starters, Teragram is a tiny firm, and its technology  is certainly adaptable to SAS’s core markets.

So there are a number of ways that SAS could work in Teragram’s higher-level natural language processing into its own offerings. Nope, as CEO Jim Goodnight told us, it doesn’t mean that because 85% of enterprise data is unstructured data, that in five years SAS would transform itself into a text mining company. Its roots are in numbers, remember.

But it could enhance a medical records system, where now the goal is to ferret out which condition and treatment codes tend to turn up most frequently within different classes of patients. With Teragram adding higher level context, a context-enhanced SAS Text Miner could elicit semantic relationships among the data it unearths, where for instance, it could correlate diagnoses with context on how those diagnoses were arrived at, or whether the diagnoses were made using consistent logic.

According to Gaurav Verma, who directs marketing for SAS’s BI products, the goal on that side of the product line is to get analyses of structured and unstructured (e.g., text) data to be sync’ed earlier in the process. Today, the analyses are typically brought together after the fact – both are conducted in parallel and united at the final report stage. In other words, you have two tools performing analyses without the assurance that they are operating on common assumptions. The goal here would be for each tool to exchange signals earlier in the process, so the search and analytic criteria could be harmonized earlier in the process.

---