LEX: Another Day on the Job
It’s not every day that you try your hand at developing a sophisticated software that can adapt to unique situations presented by each user--not unlike Apple’s Artificial Intelligence (AI) service, Siri or Amazon’s Echo App. But for some, it’s simply another day on the job.
What is it?
For the past month, a select team from Systems Development and Innovation (SDI) has been “kicking the idea around” of experimenting with some new services that Amazon has made available to the public--namely, access to the same programming interface used by the Echo App. For a fee, Amazon allows access to use and, in the process, better understand this product’s innerworkings--but with some limitations.
“Obviously Amazon isn’t gonna spend years developing Alexa and then just release the secret sauce that makes it work for everybody,” offers Walt Robins, an Associate Applications Developer. “They release basically the speed track ignition and stuff like that. But everything that it knows is still closely guarded by them.”
Even still, this access has the potential to provide significant insight about how the product operates. So naturally, the SDI developers’ curiosities could not be contained. Just what is it that makes the Echo App so successful, and could SDI conceivably develop something comparable in the future? And if so, how could this be done? The answer, the team tells me, comes from better understanding the products that already exist.
“We wanted to see what the programming looks like, what it can talk to, and what we can get it to do conceptually,” provides Altai Otgonyin, the Platform Administration Team Lead.
That is why in late May, the team came up with the idea for LEX, an early stage concept for a similar service that would serve to benefit the Kent State University community in a variety of ways. For instance, ideas have been floating around to enable LEX to provide vital information about campus, including GPS for students and faculty. Furthermore, even campus visitors could conceivably use LEX to help discover the ins and outs of campus.
Although it is still just a rough idea at this point, the team told me that LEX would likely use a natural language interface and would be adaptable, but not on its own. That is, users would have to specifically tell it what to do. In this way, no machine-learning would take place, and thus, LEX would not be considered AI, but rather it would be an interface capable of adapting based on what users command it to do.
How Does it Work?
That being said, Otgonyin shares that, as of now, “LEX is just Proof of Concept to determine how easy it is, does it work, what kind of things we need to use this new thing that Amazon rolled out.”
For those who may not be familiar with what Proof of Concept (POC) is, fear not. I wasn’t either. Essentially, it is an early stage of development in which developers take a brand new technology and see both what it can do and what uses it has for their intended audience--in this case, KSU students and faculty, and perhaps even anybody interested in KSU in general.
“LEX is just Proof of Concept to determine how easy it is, does it work, what kind of things we need to use this new thing that Amazon rolled out.”
Furthermore, Robins adds, “We would have to build all of its knowledge. So on one hand that’s good because we could customize it to only have Kent State knowledge and it would know what students mean with specific jargon and stuff like that. But at the same time, part of the POC is knowing what we could do with it, given a finite amount of development time.”
Process and Challenges
So in order to test this concept in a limited amount of time, Senior Database Analyst, Kevin Hulick, and Lead Applications Developer, Rebecca Benya, came up with the idea of importing KSU’s academic calendar into a database and searching it using voice commands. Not only did the operation work successfully, but, according to Hulick, they “found out we could do lots; but, for it to do that we would definitely want involvement not from just [Information Services].”
That is to say, the team realized that in order for the LEX project to be successful, a wide university outreach initiative would have to take place.
“If we want to make something that’s useful to a general Kent State populous, we need to know what people are actually interested in using it for,” says Robins.
And, of course, this is where data comes into play. Otherwise, how can the needs of a unique student body legitimately be met? Yet, information gathering has presented the biggest challenge to the group so far, simply because so much of it is required.
“You need a lot of data to know what 30,000 people are going to be interested in.”
In addition, if the team wants to engineer a product that responds to voice commands, that means it must be familiar with the types of questions that users might ask, including all of the different variations of how to ask those questions.
Moreover, once data is gathered informing what types of information or services users are interested in using LEX for, the team must consider how LEX will answer those questions. Because of this, the team tells me that the front end interface for this project (including all of code that makes up the visual appeal of the product) will be harder to build than the back end (e.g., the code that helps the product function logistically).
Again, in order to meet this challenge, Robins adds that SDI “must gather data from the Kent State populous about what sorts of things we want to assist in.”
Currently, the POC stage is over, which means that if the LEX project is determined worthwhile, then the developers will move beyond this stage. Specifically, SDI will begin to seek involvement from other departments on campus regarding how to incorporate data on what student needs are and how this service could meet those needs.
Still, Otgonyin stresses that, “This is still in the very, very early concept stage. There hasn’t event been a decision made on whether or not this is gonna proceed into a project. So there’s a chance that nothing could come of this, but there’s also a chance that this could develop into something as well.”
But either way, it’s plain and simple that Systems Development and Innovation is seeking new and innovative ways to develop new products and services to better aid the technologically ever-progressive KSU populous.
Be sure to watch out for FlashLine notifications in the future to keep up to date on new projects and developments!