People attending the Bio-IT World Conference and Expo held in Boston in April were wowed when a presenter used an iPhone to execute a molecular modeling simulation on the cloud with a few simple voice commands. That presenter was Bas Burger, president of global commerce for BT Global Services, and the clever minds behind the stunt were the people who make up BioTeam, a small company that provides informatics solutions to life science researchers.
The demonstration used Apple's Siri software, a natural language processing application that comes standard on the iPhone 4S and allows users to operate their iPhones with just their voices.
Burger first asked Siri to run the NAMD molecular modeling software application on his company's cloud service, BT Cloud Compute, to which Siri responded: "It depends — is this Bas?" After he confirmed his identity, Siri said, "Hi Bas. Yes, this experiment has been approved to be within your budget." Burger next told Siri to "please run the experiment on BT Compute," to which Siri replied, "OK, I have started the experiment for you." He then asked Siri to confirm the status of the experiment and return the results, which Siri then displayed on the iPhone screen.
Not quite there
If it seems too good to be true, that's because it is — at least for now. The application Burger demonstrated is not yet publicly available, and the presentation was intended to drum up excitement around the telecom giant's new compute cloud aimed at the life sciences market, which is called BT for Life Sciences R&D.
"This idea came during a BT 'hothouse' session that we ran with customers, partners, [and] internal and external consultants during the validation stage of our BT Life Sciences R&D proposition," Burger says. "We were investigating options for natural language recognition interface versus traditional interfaces during this hothouse session. The teams came up with this approach and then mocked it up and presented it during the final presentation to the judges' panel. Once iPhone 4S devices became available, we felt that it would make it a natural choice for testing natural language interface for cloud approaches due to its pervasiveness as a consumer device of choice."
[ pagebreak ]
The Siri demo was built on a hack engineered by BioTeam founding partner and principal investigator William Van Etten. Siri normally works by sending a recording of the question or command to servers at Apple, which then statistically analyze the sound wave to make sense out of it so that Siri can provide an appropriate response. Van Etten's implementation was specifically designed for the life science user, who uses a specific vocabulary. The vocal commands are rerouted to a proxy server hosted by BT Compute that screens the commands for matches to a list of phrases related to the workflow analysis pipeline. If the proxy server detects none of the key phrases, it sends the voice recording along to Apple's servers for normal operation.
Developers at BioTeam and BT Compute say this was no dog-and-pony show and that they are serious about the future of natural language processing in bioinformatics. In fact, they say voice-controlled computing is the next logical next step in how people interact with computers — and not just as a universal access tool for the disabled or a faster way to run a program than typing or clicking.
"As today's sequencing technology moves from laboratory instruments to devices and diagnostics areas, you could imagine a patient in a hospital or at a doctor's office where the attending physician uses Siri communication to order a full set of tests, including complete genome sequencing, identification of associated biomarkers, and even finding donors for bone marrow transfers by contacting national bone marrow databases," Burger says.
Officially, Siri has only existed in a beta version, and Apple customers have criticized the app for its frequent mistakes, both in the interpretation of the voice commands by the user and the responses provided. However, part of natural language processing is its capacity to learn and become more accurate. Van Etten says every time he uses Siri, it seems to get better, which is a result of Apple's severs and statistical analysis software adapting to users' language patterns.
Talking HPC
Enabling researchers or clinicians to use voice commands to run experiments on the cloud or to access databases is not just about ushering along the next evolutionary step in computing history. Instead, Van Etten sees natural language processing and voice-controlled computing as a way to lower the barrier to entry for high-performance computing for the whole community.
"I remember 10 years ago when the only way we could perform data analysis in the life sciences was through the Unix command-line interface," Van Etten says. "Now there's been a lot of effort over the years to put a Web interface to those applications, but there's still a bunch of drop-down menus, check boxes, dials you can turn, and if you don't understand the depth of what they all mean, it's hard to approach. But, if you can have a conversation with an expert in the form of a computer interface, that's a totally different experience."
[ pagebreak ]
Natural language processing may enable improvement to computing, and anything that makes computing easier for researchers is going to be the focus of the BioTeam's energies, Van Etten adds. While he and his team will continue to work with Siri once Apple releases its Application Programming Interface development kit for the technology, they are not waiting for the release to further their natural language processing efforts. Within the last year, BioTeam has started working with a number of biotechnology companies to develop prototype solutions based on natural language processing that integrate the companies' own internal applications and databases.
While natural language processing has not yet seen widespread adoption within the bioinformatics community, the marriage of mobile app technology and genomics is certainly gaining momentum. In April, Illumina launched its own version of Apple's App Store for bioinformatics software called Basespace Apps, from which users can download and install a range of bioinformatics software for next-generation sequencing analysis. That same month, the company released its MyGenome iPad app, which allows users to access their genetic data from that device.
Burger says his company will continue to explore the intersection of mobile apps, cloud computing, and natural language processing for possible research solutions, though he notes that there are a few kinks to work out. "First, we need to get the fundamentals right — which are potentially more important at this point — including an app store design, adding additional open-source and commercial data sources and applications to our environment, ensuring security and compliance are fully addressed, maximizing throughput, developing collaboration options for accelerated decision making, lab instrument and devices communication with our cloud environment," he says. "A natural language interface is just one of many things that we need to do in the coming months to ensure our platform is best suited for life sciences research and development."
Those who focus on the mobile aspect of this demonstration — while impressive — are missing the point, Van Etten says. "When people see Siri, they think of this as a mobile application, but it's just a mobile use case," he says. "There is a great deal of computing that will be done by communication with a computer through voice, and in many cases it won't be so much that the human is the expert — we're going to find places where the computer is the expert. Whether it's this year, five years, or 10 years, what you've seen on Star Trek will be how we get our computing done eventually."