Virtual Personal Assistant Sherlock Aims To Make Human-Computer Communication Elementary

Virtual Personal Assistant Sherlock Aims To Make Human-Computer Communication Elementary

Researchers at IBM and Cardiff University have developed a new virtual personal assistant called Sherlock. The software uses information from users to make human-computer communication more efficient.
(Photo : Cardiff University)

Just as the iconic character Sherlock Holmes used the art of deduction to solve the most enigmatic cases, a new digital assistant with the same name as the famous detective developed by scientists at IBM and Cardiff University tries to make it so there's no room for mystery in the communication between the software and its users.

The new virtual personal assistant known as Sherlock (Simple Human Experiment Regarding Locally Observed Collective Knowledge) works with smartphones and tablets, similar to Apple's Siri, Microsoft's Cortana or Amazon's Alexa. Sherlock uses a "controlled natural language" that is said to better understand how humans think and talk and was developed by IBM, which has famously been researching the capabilities of artificial intelligence for years through its Watson program.

Instead of learning to search the Internet or databases for answers to your questions, Sherlock uses a "local knowledge base" that it has built up based on what users tell it, which helps it acquire new information and act upon it.

"With Siri I take it out to ask for directions and it gives me driving instructions even though I always take the train," project leader Alun Preece of Cardiff University's School of Computer Science & Informatics told the BBC. "I would tell Sherlock that I take the train and so it would know that I always take the train and would give me that information."

Sherlock can also use crowdsourced information to help festivalgoers stay away from the busiest areas or help medical teams or aerial drones find people in need of aid during an emergency, for instance. It receives this information by pinging users with a request for information based on another user's question. They then text their best approximation back to Sherlock, which will process all of that information and send a "best guess" back to the original user via text message, according to the Wall Street Journal's Digits blog.

Sherlock currently interacts with users through text, but there are experimental versions of the software that use Google's Voice interface and Apple's dictation interface on the Apple Watch.

Sherlock doesn't yet "have a particular use or niche," and when it does, it will only be able to respond to specific tasks, Preece told the BBC. We'll just have to wait and see what problems this software will potentially solve.