Semantic search – an Interview with Brooke Aker
For my second expert interview on the semantic web I set out to find a key commentator who is currently involved in the heart commercial semantic search. Can someone like that describe how these web developments will impact on coal-face journalists and researchers? I didn’t need to look far. Brooke Aker is an expert in competitive intelligence and before taking up his post at Expert System he formed both Acuity Software and Cipher Systems. He has worked with 130 of the Global 2000 in the formation and operation of successful intelligence and is a key commentator on the semantic web.
Expert System is a leading provider of semantic software which discovers, classifies and interprets text information. Its semantic software, Cogito, has been deployed across most industry sectors and the company’s clients include Eni Group, Pirelli, Microsoft and Telecom Italia.
Social networks and semantic search…
Q: Do you think the explosion in growth of niche sites such as Xing and Peer Trainer will accelerate the demand for semantic-type applications that allow people to travel seamlessly through various social networking services?
Brooke Aker: I would agree with this. Facebook and Myspace are good examples of getting people used to the idea that they can, not just search, but connect to people and content. So they set the stage for users to migrate quickly to Web 3.0 properties where users can search for and connect, analyze, and assemble very specific people and document objects in ways that are uniquely designed by them.
Transforming the way you work…
Q: You recently released a (very useful) presentation on ‘what is semantic search‘. It remains, however, difficult for coal-face researchers (insite readers!) to grasp its significance. What, in your view, are the best examples of semantic search that hold the most promise. I’m thinking here of apps like Juice. Not tools that help publishers – but tools that can currently help people in their day to day work?
Brooke Aker: We have been involved with applications that use semantics and combine search with discovery, or search with analysis. Let me explain….
Because semantics expands and connects similar concepts, from where I begin my search, I may end up in a place I did not expect. Say I run a search for “stock” and ask to limit my search to the concept of stock in the sense of soup. This helps avoid stock as in “inventory” or stock as in “equities.” Now, I tell the system to expand the concept of soup stock and I get bouillon, stock, base, and a completely new word to me called fumet. I can then reduce my search results further by noting Emeril Lagasse is mentioned as a chef in one of the documents extracted. So in the end, I used semantics to search for a recipe on soup stock and ended up in a precise but completely new place with a recipe called “Emeril Lagasse’s classic fish fumet.” The document had no mention of the word soup or stock. This is something I would have otherwise missed.
For search combined with analysis, we often will employ semantics in a modeling sense. Think about a competitor who may be preparing to launch a new product, but the company has not made anything public yet. We know what steps that company must be taking in order to launch a new product: things like ramping up production lines, buying new machines, contracting with ad agencies, hiring new people with specific skills, etc. These actions are likely to be public. Semantics are employed to broadly find these indicators, which feed a model. If enough of the indicators are present, the model concludes a new product is forthcoming. So here, semantics plays a predictive role. Such foreknowledge of such things is many times more valuable than simply knowing the moment something is reported in the press as having already occurred.
Tech stacks…
Q: Many people, I think, assume that the semantic web will usher in a new period of improved search. But, in fact, developments such as the ‘social semantic desktop’ like Nepomuk may accelerate the development of semantic web technology. Do you agree?
Brooke Aker: I agree. The ideal architecture would be to re-index the entire Web semantically and have new browsers to read it. But that seems like a long shot for the time being. So instead, if you embed the semantic processing of every html page the standard browser reads, stores and retrieves locally, you have in effect federated the problem across the Internet. And of course those same special browsers or browser plug-ins could also peer-to-peer share their semantic results if directed by the end user.
Filter failure…
Q: I really like your graph visualising the downside of web2.0 (in your presentation) – as more and more information is mass produced there is a danger that productivity may slide. Do you agree with Clay Shirky’s recent argument that ‘information overload is just filter failure?’ What we’ll all have to get used to is using the right filters at the right time and learn how to maintain them?
Brooke Aker: Filters are a blunt instrument to a more delicate problem. And it implies a lot more work on the users end. This spells failure to me. People want convenience and simplicity, and are already overwhelmed.
About Cogito….
Q: Can you describe some of the practical applications to which your software, Cogito, has been applied (and therefore, where Expert System USA is positioning itself)?
Brooke Aker: One of the best examples we have now is in customer service. We support the online help function of mobile devices using a natural language interface. So users type a question they have about how to operate their devices into the handset. We return 1 precise answer to them. This prevents the user from doing two things. First, they don’t return the device having found it too difficult to operate. This also saves the company the cost of acquiring the customer before they can earn it back. The second thing it does is deflect an inbound call to the customer service center where the average cost is $20. We can give the direct, accurate answer for about a ½ a penny.
Q: In your presentation you draw a green figure that demonstrates how web3.0 will allow people to better ‘filter’ or pick the content that is relevant to them. In your view, what single application/product best demonstrates the power of this technology?
Brooke Aker: Yes, this is the example I gave about the fish fumet. The product we have made to do this is Cogito Focus. It is a corporate search tool that includes crawlers, semantic indexers and a nicely done interface that is not much of a stretch over a conventional search box interface (e.g. little training).
Viable applications….
Q: I was interested to read the “On the cusp” by David Provost. In it he concludes that companies are on the verge of constructing very practical and commercially viable semantic applications. (Provost makes the point that Twine is succeeding because it has ditched semantic terminology and focused on the ‘business mission’. While the terminology isn’t obvious – the semantics under the hood is). Do you agree?
Brooke Aker: Sure I do. Business users don’t care how it works, only that it does work and provides some visible, measurable value. Same is true for consumer-facing applications as well. Can I do something helpful and valuable that I could not do before? That’s what matters.
Tags: ad agencies, application/product, Brooke Aker, Cipher Systems, Clay Shirky, combine search, consumer-facing applications, corporate search tool, David Provost, Eni Group, ENI SpA, Expert System, html, improved search, Microsoft, Microsoft Corporation, online help function, peer-to-peer, Pirelli, Pirelli & C. Accomandita per Azioni, semantic applications, semantic processing, Semantic search, semantic software, semantic technology, Semantic web, semantic web technology, semantic-type applications, social networking services, Telecom Italia, Telecom Italia SpA, USD
As outlined in my 
Not many people are as close to the heart of the Semantic Web as John Breslin. John is the founder of the 
