FAQ

My keyword search results weren't as good as Google's - why?

Belief is fantastic at discovering concepts in unstructured text; text like you'd naturally write or speak in sentences. Keyword searching has its place, but consider that expressing just the right keywords is sometimes hard when you want to communicate concepts unambiguously. What does "red conductor" mean? Musician from behind the Iron Curtain? Or a copper wire sheathed in red insulation?

So if not keywords, then what?

Try more text. Either a paragraph of text, or a full URL.

Is that useful?

You bet it is. If you are reading a web page on an issue of the day that really interests you, BeliefNetworks can find more articles like it. That's because BeliefNetworks can understand the concepts of that article. It's like going to the library and showing the librarian a particular book and asking him if he has any more like it.

Do you have a demo of any of this?

Yes, an excellent one at http://mashmeup.com. Start with a web page, and either copy a selection within it, or copy the URL of the page, and paste that into the "Search Phrase or Site" field. You will be returned three lists of resources that relate conceptually with the page you started at: related web pages, related photographs on Flickr, and (this one is neat) related live-tweets ongoing on Twitter!

"Relating conceptually" - what does that mean?

It means that BeliefNetworks has semantically analyzed your starting page or text for conceptual information; in other words it "understood" your page. Likewise, it has semantically analyzed all the target resources, like web pages or the Twitterverse, and distilled their meaning and relevant concepts. After that it's a matter of matching concepts in your page to concepts from the target resources and ranking them for relevance.

Live tweets from Twitter? Is that right?

Sure is. We are white listed to monitor public Twitter conversations. Every tweet is conceptually laden, enough to communicate information to human readers. So, BeliefNetworks can mine the tweets for concepts as well. As soon as a tweet is analyzed, its conceptual information is ready for relevance matching.

Is it fast?

Yes. While there may be a lag in receiving information from Twitter, once that information is received, BeliefNetworks extracts the results a lot more quickly than it took to receive them.

Does that mean that BeliefNetworks is a Twitter application?

No. MashMeUp is a "Twitter" application per se, but it's also a demo of the functionality that BeliefNetworks makes possible. BeliefNetworks' services can act against any live data feed.

So how must the data be formatted?

Data analyzed by BeliefNetworks does not require an a priori formatting. BeliefNetworks can also analyze historical data and lots of it. The data can be unstructured (read: free text) or semi-structured (like XML) or structured (like a database).

So what would it take to integrate my custom web application with BeliefNetworks services?

You can write your application against the BeliefNetworks Public APIs, for one. BeliefNetworks allows up to two thousand queries per day against RESTful web services. You can find information on web pages, Flickr and in the Twitterverse, just like MashMeUp does.

Writing an application using the same resources available to MashMeUp seems interesting, but what if I need to analyze my own information? What if I want to do relevance matching against my own business data?

BeliefNetworks supplies "custom" APIs to allow you to build repositories of analyzed information ("corpora") from your own data. The same daily limitations apply, but they are generous enough to prototype and also suffice for the needs of many applications.

What's the story with Ted and doughnuts?

CTO and co-founder, Ted Tanner, has never met a doughnut he didn't like.