rsz_app4AppCrawlr, the semantic mobile app discovery and recommendation service, is adding Windows Phone apps to its iOS and Android app-finder solution. The service, which debuted earlier this year (see our story here), has at its heart the TipSense content discovery and knowledge extraction technology, which also lies behind the company’s DishTip food discovery service.

Microsoft has been trying to pick up some steam in the mobile phone market, acquiring Nokia’s phone business for $7.2 billion in September and this month announcing Update 3 of Windows Phone 8 that supports faster processors and larger screens. Its market share has grown, according to Gartner, to 3.3 percent worldwide, to put it in the #3 smartphone slot. Gartner also noted, though, that Microsoft has work to do to grow interest among app developers.

Microsoft recently calculated more than 100,000 apps in the Windows app store. Though that’s far less than what you can find in the iPhone App Store – which reportedly boasts something along the lines of 1,000,000 – it’s still nice to have some help to find what you really want among tens of thousands of apps. From every app detail page, users can search on a single platform or to find across Windows Phone, iOS, and Android operating systems.

The main reason the company added Windows Phone Apps to AppCrawlr is because it had numerous user requests for the feature, says TipSense co-founder Joel Fisher. “Windows Phone users want a better app discovery and recommendation experience and were frustrated that only iOS and Android users could find apps on AppCrawlr.  So, even though there are fewer Windows Phone users vs. iOS and Android, they still need a way to find the best apps,” he says. “And, remember…even though Windows Phone may not have millions of apps today, they still have a sizeable amount.”

Whatever OS platform is behind your smartphone, the TipSense technology is constantly evolving to improve the discovery process, says co-founder David Schorr. The advanced machine learning system, which surfaces the latent information in unstructured text content to convert into actionable intelligence, can systematically learn the underlying concepts and relationships for any given product or content area, he says, and then leverage that knowledge graph along with content fingerprinting to enhance recommendation, functional search, app discovery, diversity of results, predictive analytics, and more. In addition, AppCrawlr statistically infers user profiles and motivations based on app usage down to a very granular functional and feature level to enable personalized search.

Taken together, Schorr says, “we can drive the app discovery market forward through a host of innovative and differentiating features.” That includes uncovering various aspects of an app (such as replay value or value of in-app-purchases) and then using that information to better tailor results. “They can allow for guided discovery which helps users better refine their intentions and navigate the millions of apps available,” he says. “They can also find more apps based on the ones they already like.”

In AppCrawlr, TipSense’s deep data analysis of app content creates groupings covering every niche of apps, helping users find out about apps they might not otherwise easily come across – hack-and-slash games, for instance. (Hey, Halloween isn’t that far gone.) As Schorr explains it, TipSense realized that there was a cluster of games around the concept of hack-n-slash. “From statistical analysis, it built a content fingerprint and knowledge graph from tens of thousands of aspects that it uncovered,” he says. For example, TipSense concluded that hack-n-slash is a type of shooting game with functional similarities to combat games and fighter games, with fast-paced actions, with popular features of character customization, upgrading weapons, and multiplayer, and that are often popular with RPG fans and casual gamers.

“Instead of manually creating a taxonomy to represent the app space, TipSense allowed us to automatically generate a very rich knowledge graph,” he says. “We dynamically built a domain meta model from actual user content, which typically offers much more relevant results since it reflects a ‘cultural understanding’ of how people interact with an app in the real world.”

AppCrawlr is good for users, and good for app developers and ad networks, too, he says, allowing them to gain deep insights into a variety of user segments and other aspects of apps to help improve conversion rates and campaign performance.

The TipSense technology can be applied beyond its current food discovery and app discovery incarnations – travel, movies, TV, books and so on. Though Schorr isn’t specifying where it goes next, he believes that there is no shortage of business applications that can benefit from actionable data that must be extracted and surfaced from unstructured content. “The companies that can best understand all the buried signals, and turn that data into actionable information will have a tremendous advantage for offering better content, personalization, recommendation and other types of information,” he says.