Personalizing computer architectures
First there was personal computing; now ECE researchers led by JoAnn Paul are exploring personalized computing — how computer architectures can be customized to individual usage patterns.
“The relationship between users and computers has changed dramatically in recent years, due to the combination of wireless communications, smaller computer sizes, longer battery life and the ubiquity of Web pages,” Paul says. “At the same time content – particularly Web content — has changed and user access to it has become more individualized. The need to more efficiently access the Internet from ever-smaller computing devices suggests that the time has come for personalized computer architectures, based on the way individuals use their computers.”
The research team envisions devices that are more responsive to user needs and are faster while using less power. “A lawyer who follows baseball could use a different cell phone architecture from that of a stock broker who enjoys travel and fine dining, just based on the way each would interact with their device (language processing) and the dominant websites they follow,” she explains.
Most people use their computers for different reasons at different times, the most dominant separation being that between work and play, she says. “We are initially investigating how to optimize language recognition for those specific contexts,” she explains.
With a nod to chemical analysis, her team has developed what they call a spectroprocessor. “Instead of identifying the composition of matter by sorting according to concentrations across a set of elements, we identify the meaning of data by sorting according to its relevance across a set of user interests,” she says.
In a first step, her team is developing recognition algorithms for search queries that work with personalized contexts. “The content of a query and the context in which that content is searched, is strongly related to the user,” she says. “If we can develop search methods that cut through all the data depending on context and user, we can extend that to personalized architectures.”