Computing with rich context – HP Labs China expands the frontiers of context-aware computing

Context-aware computing isn’t new, notes Min Wang, director of HP Labs China. “But it really has a new meaning now,” she says.

Typically, context-aware computing has encompassed operations, like search, that take into account a few simple parameters such as location and time of day. Thus a smartphone search for ‘pizza restaurant’ might prioritize results for places both nearby and currently open.

“But today,” says Wang, “we have information available to us in such a volume and in so many dimensions, that we’re facing a huge challenge in how we sort it out.”

That leads researchers in HP Labs China to explore more sophisticated contexts that could help us make the most of the information presented to us in a wide variety of computing situations.

They’ve created a coupon recommendation application, for example, that goes beyond the parameters of key word search and location to offer coupons also based on a user’s web browsing history and social networking activity. It might offer you a coupon for a specific salsa available at a local store, for instance, if you’ve tweeted about your love of Mexican food and also recently searched for a variety of taco recipes.

Coupon Recommendation

A second project extends the utility of HP’s Smart Print tool. Smart Print selects the areas of a webpage that you’re most likely to want to print, leaving out less relevant framing material that few people ever need.

“We found that Smart Print did a good job, but could do better,” says Wang’s colleague Shimin Chen, who is the research manager of the Data Management and Information Analytics (DMIA) team at HP Labs China. For one thing, Smart Print requires you to manually select your print area before generating a final print document.

“What we’ve done,” explains Chen, “is take advantage of other people’s previous print selections for pages with similar structures, which are saved in a print log database. These help us offer print selections to new users that are more likely to be perfect first time.”

Improving Smart Print with Crowd Intelligence

Context in the enterprise

Both of the above studies exploit the context related to the web or social media, but many other contexts can be used to improve system or service performance.

The enterprise environment, for example, typically features rich stores of both structured and unstructured data that can be drawn upon to improve IT services.

To demonstrate this, Chen and his colleagues have collaborated with HP’s Software BTO team in Shanghai to create an application that can better resolve new IT support queries. It does this by intelligently reviewing past support requests (held as unstructured data) and finding solutions previously offered to people with similar IT equipment (held as structured data).

Another enterprise collaboration, this time with HP security subsidiary ArcSight, explores how ArcSight’s popular Logger product can better serve non-specialist or first-time users. Logger offers a highly sophisticated query mechanism, but that also makes it complex to use. By adding a contextual layer of previous queries to Logger’s search functionality, the HP team has been able to automatically help new users search more effectively.

The operating condition of a network running any specific IT service can itself be a source of valuable contextual data.

In another collaboration, this time with HP Networking, Junqing Xie, who is the research manager of the Networking group at HP Labs China, has been working with his team to leverage context information to make operations (e.g., HTTP compression, web acceleration etc.) in network systems (such as Application Delivery Controller) more efficient. In such case, context information of the system, the network and the application/user delivering the traffic will be taken into consideration.

“Traditionally, we’ve done this without much awareness of context,” says Xie. “In our work, though, we’re trying to leverage not only the content of a web page, but also the conditions of the system (e.g., CPU load and memory consumption), as well as the conditions of the network through which we are delivering it – such as the bandwidth available to the user, the kind of device the user is employing. In that way we can judge which web objects we’d like to compress and cache – and make them available to other users in similar situations, as well.”

A context-aware HTTP compression framework in ADC

Towards a framework for next-generation context-aware computing

All of the above projects are application-specific, notes Wang.

“But we’re now also looking to figure out the commonalities that exist between these studies,” she says, “and to ask what it would take to build a general purpose framework for context-aware computing that takes advantage of existing cloud and database techniques.”

“With the information explosion on the Web, within enterprises, and in network systems,” Wang adds, “this will only get more important. In the future, asking how we can adapt application or system behaviors to the relevant contexts of users or systems will be the key to achieving user satisfaction and high system performance.”

Context Analytics Framework