The semantic technology platform behind restaurant dish discovery service Dishtip (which The Semantic Web Blog discussed here) has made its way to a new domain: mobile apps. The company last week unveiled AppCrawlr, which uses its TipSense content discovery and knowledge extraction technology to cut through the noise to help users find the app that’s right for them in a world of hundreds of thousands of options for iPhone, iPad and Android devices.

“With traditional search models there’s no easy way for guided discovery to narrow down from all the apps out there to what you want,” says Dave Schorr, who with Joel Fisher is a co-founder of TipSense LLC. Keyword searches aren’t going to help you find apps that help when you are having a bad day, for instance, or understand that someone looking for a dating app (as in relationships) is looking for something different than someone looking for a date (as in scheduling and productivity) app. But searches on AppCrawlr can suss those out, taking data from from all across the web – blogs, tweets, reviews, and so on – and surfacing and organizing the concepts and topics buried in all that unstructured data.

“It’s a new paradigm to manage a large data set,” says Schorr. “We’re using concepts to come up with a much better experience for discovery.”

The TipSense semantic platform that drives Dishtip has been continually enhanced to boost the actionable signal from noise ratio since that service debuted in 2011. AppCrawlr uses a slightly different corpus and domains, but a lot of the underlying conceptual entity recognition is the same. The major new feature that really comes into focus for AppCrawlr, because of the hundreds of thousands of different intentions that can come into play in app searches is content fingerprinting.  “We are able to create a uniquely Identifiable content fingerprint for pretty much any content, topic and domain across any corpus,” Schorr says. “That is the key enabling piece for us to get a higher quality of signal and do a lot of this type of discovery and better search.”

Content fingerprinting and its word-sense disambiguation is what gives the co-founders the confidence that dating and date can’t stem together, because their fingerprints are so different. Look at it this way: Someone who likes Angry Birds might be looking for an app like it, but what’s like it goes beyond just the fact that it’s a game, that there are ratings for levels completed and so on – the things that traditional recommendation system might recognize.  With content fingerprinting, Schorr says, you have a much more fine-grained set of features to work with for recommendations – addictiveness, replay value, graphics, sound effects and more. Using those conceptual-based concepts, AppCrawlr can provide more relevant and interesting results to users, he says, and explain those recommendations to the user with an overview of the key concepts, alternatives, comparison charts of concept rankings for products in the category, related recommendations, and more.

That’s unique to AppCrawlr, Fisher says – and it’s a uniqueness that others leveraging the TipSense technology can take advantage of, he thinks. Apple and Google, for instance, can’t show their users all this information on their app store sites, but if they did, they might be able to add some more revenue to their coffers. “If some company like Apple were to use the TipSense technology, it could provide the iTunes store a much better shopping experience, where people would find the apps they are looking for and Apple would be able to better up-sell them based on their interests in certain features and how important those are,” Fisher says. TipSense brings to light both free and paid apps with its feature by feature attribute comparisons, and a user viewing that mined data – information that might otherwise get lost in the noise – might decide that it’s worth it to spend a buck or two to buy the productivity app that stands out for its organizational tools rather than the free one where those tools aren’t rated so highly.

Netflix, Amazon, TripAdvisor, Facebook – TipSense founders see opportunities for them all to add value with a little TipSense technology. Fingerprinting, Fisher says, works as well in its tests for news and Q&A as it does for mobile app discovery. “We see and read, for instance, how Facebook keeps talking about figuring out how to provide answers to questions people have and make use of all the user content they are sitting on,” says Fisher. Witness last week’s debut of Graph Search.  “If you look at the social graph, the challenge is the sparsity of data. Just mining at the top-level features, it’s a hard problem to solve. The data isn’t there. We see TipSense technology kind of filling in those gaps, to provide a rich set of concept-based features to fill in all the missing areas in the social graph.”

Imagine wanting to discover all the food your Facebook friends like that has some relationship to Taco Bell cuisine: “From the concept model you’d get all the concepts related to Taco Bell – that it’s a drive- through, open late at night, maybe an age demographic. All those concepts and topics are very clear and those are the extra features to make social graph search successful,” Fisher says.

One thing the co-founders have adhered to both in Dishtip and AppCrawlr is making the sites easy to navigate and understand for the average user. The downside, Fisher says, is that people may not understand everything that’s behind that interface. “When people see this, they don’t feel like this is machine intelligence, like this is leading edge semantic technology,” he says. That’s good for getting them onboard but he hopes it won’t keep the market from realizing the power of the underlying technology. “Just because it doesn’t look like a science experiment, it may not sink in that what they’re looking at something really advanced.”