Dataspaces poised to be next big thing from Google
From time to time we check in with search analyst Steve Arnold. Over much of the past few years, he has spent quite a bit of time analyzing Google. Particularly, he looks at what patents the company files to help understand the company's closely held business plans.
Recently Arnold published a book, titled "Google, The Digital Gutenberg," which provides an overview of how all of Google's various offerings, such as dedicated YouTube channels and Google Books, work together like one gigantic publishing system.
One of the interesting things Arnold has unearthed is a new Google undertaking called dataspaces. "Dataspaces are a way to manage information and make it possible to run certain types of queries impractical in traditional indexes and databases," he writes. "Dataspaces are constructs that integrate many separate indexes, their metatags and data."
Dataspaces tackles the age-old problem of making sense of unstructured data — all the Web pages, e-mail, word-processing documents and anything else that hangs off the Internet but isn't filed into a properly architected, structured database. Today, Google already makes a mint from indexing such unstructured data and serving the results up in search queries. But if the company can find some way of making all this stuff machine-processable, its services can become even more valuable to people, or so Arnold's theory goes. This is the work of dataspaces.
Where did dataspaces come from? In 2005, Google acquired a small company called Transformic Inc. which had technology that could merge structured and unstructured material into a single repository. There are plenty of companies with software that could do this sort of thing, so Transformic must have some sort of special sauce for it to be gobbled up by Google.
Tarnsformic's technology was largely developed by Dr. Alon Halevy, largely while at Bell Labs, according to Arnold. Halevy focused on how to extend database-quality queries into ambiguously defined data using lineage, statistical likelihoods and a number of other fuzzy-logic-like reasoning. A dataspace system can not only process ambiguous data, but also incorporate new information as a search increases its scope and reconcile conflicting sets of data to derive more definitive information.
If such a system were to work, it could change the enterprise software market profoundly, Arnold asserts. "The challenges of petabyte data sets have hamstrung many organizations trying to make traditional work processes and relational databases work in a cost-effective way. Google could use its existing infrastructure to perform automatic transformation, advanced processes based on smart software, and introduce new types of queries to users."
In short, the dataspace invention makes possible new types of auto-generated outputs because the system can deal with uncertainty and varied lineage of information," he argues. Such queries could say build complete dossiers on people or organizations, which would certainly be of interest to intelligence and law enforcement agencies. Investors could use the technology for ranking investment advisers. Consumers could use it to build a better, more customized profiles of some product of interest.
Of course, Arnold is making a lot of leaps of faith here. Many smart people have been trying to tackle the problem of making sense of unstructured data, with middling results thus far. And even if such a sorting technology could work, it is another question entirely if it could be ramped up to the scale needed to serve millions of people on a daily basis. In the book, Arnold doesn't go into the specifics of how this technology works or what makes it better from everything else on the market. And even if it does work, whether Google will have success with it is far from given.
Nonetheless, "Google: The Digital Gutenberg" is a good primer for understanding the direction that the search giant may be taking, which is especially valuable given how tight Google itself is with sharing such information. For a deeper dive, you may wish to check out a talk that Arnold will be hosting in a few weeks.
On Sept. 23, Arnold, along with Arpan Patel, director of Somat's Information Engineering practice, will explain dataspaces technology in more detail, a briefing " Change 2010: Responding to Real Time Information, Open Systems and the Obama IT Vision," to be held at the National Press Club in Washington. Robert Steele, chief executive officer of OSS Inc. and founder of the Marine Corps Intelligence Center, and Jim Orris, director of Adhere Solutions, the Google partner responsible for U.S. federal government sales, will also speak.
Posted by Joab Jackson on Aug 31, 2009 at 7:05 PM