Exago Collaborates with RPI on Natural Language Processing Engine
Kingston, NY - Exago Inc. has partnered with the Rensselaer Center for Open Source (RCOS), a computer science research center at Rensselaer Polytechnic Institute, to build a Natural Language Processing (NLP) engine for business analytics. Last week, a handful of students involved in the project visited Exago’s Kingston office to showcase their work and plot a course for future development.
NLP, the ability of a computer program to understand spoken language as popularized by voice assistants like Siri and Alexa, has the potential to revolutionize how casual users interact with business analytics applications, making it even easier for decision-makers to access and manipulate company data. By allowing users to query databases using conversational speech, NLP could dramatically improve BI adoption rates.
Exago’s partnership with RPI began late last year when David Goldschmidt, Ph.D., CS professor and RCOS Assistant Director, invited Exago CTO Stewart Meyers and VP Business Development and Technology Scott Epter to pitch the NLP project to an auditorium full of RCOS students, mostly in computer science. RCOS, a program most describe as somewhere between a class and a club, gives students the opportunity to begin their own open source projects and chaperones them through the multi-semester process of seeing those projects through to completion.
“Although RCOS is overseen and advised by faculty, students are certainly the driving force behind its success,” says Goldschmidt, who works with Wes Turner, RCOS Director, to run the program. “Student mentors are each assigned a few projects, with student lead coordinators orchestrating much of everything else. Our role as faculty boils down to meeting with lead coordinators to help obtain resources they may need, handling the administrative duties, and coordinating with outside companies and other faculty doing work applicable to the world of open source software.”
Meyers had no guarantee that his pitch would land. “Not all the projects get started,” he explains. “It really depends on what the students are interested in.” But a half dozen students expressed an interest in realizing Meyers’s vision, and the project took off.
“I've always been interested in financial technology,” says John Gebbia (’20), a sophomore concentrating in Data & AI with a minor in Economics. “So this is exactly what I was looking for.”
Gebbia, along with teammates Subin Samuel (‘20) and Gabriel Vande Hei (‘20), led the prototype demonstration at Exago’s Kingston location last Tuesday. After researching and analyzing currently available technologies, the team settled on a system that uses Google’s Dialogflow language-recognition interface to map user input to an “intent,” which is then fulfilled by a custom webhook, querying a database if requested and then passing the response back to the user. The prototype uses Google Firebase, but Gebbia and his team say their next step is to adapt the webhook so that it can query relational databases in SQL.
"The difference between this class and other classes is that they want you to fail this one so that you learn a little bit."
Meyers called the demonstration a success, praising the proposed solution as “more developed” than others he has seen. He and Exago’s development team look forward to watching the solution evolve.
“The fact that we’re working with the students directly means that we're influencing the direction of the project,” he says, “and it also opens up a path for recruiting. We've had great success with students from RPI. They get really high quality kids and work them hard.”
For the students, RCOS actually serves as a sort of reprieve from the rigors of more traditional academia. Vande Hei says the class has “no actual structure” and can be summed up as “peers helping peers create cool things.”
RCOS’s emphasis on autonomous decision-making was a paradigm shift for Samuel, who says he has never been in a class that did not present him with explicit instructions and guidelines. “The difference between this class and other classes is that they want you to fail this one so that you learn a little bit. In other classes, if you fail, you fail.”
All three team members agree that there is a benefit to working on real-world projects for actual software companies. Not only do they appreciate getting a look behind the scenes at how professional development teams function, but they also enjoy working on technology that they know will have an impact.
Gebbia, Samuel, and Vande Hei all plan to continue working on the project with fellow students Nicholas Nazario (‘20) and Jiongzhen Chen (‘21) over the summer and into the Fall semester, consulting with Meyers and Goldschmidt as needed.