We talk to our computer by asking it a question and then the computer talks back, trying its best to answer it. At least that’s how the conversation generally happens between humans and A.I. Sometimes we’re happy with the computer’s answer and other times, not so much, and we either try again until we’re successful or give up. But real learning takes place with true dialogue, when there are successive exchanges that deepen understanding and where either a person or a computer can start the conversation. For people and computers to learn, good data is very important — as are ways to access it quickly. Even more important is constructive communication driven by language. Corey Patton is the Co-Founder and CEO of Pramana Labs, and he thinks prose and narrative conversations are the future in communication between A.I. and people.
“We created a way to learn about a relational database using training processes and NLP models that allows a user to just ask the question in free text. What are the most home runs that any Angels outfielder has had in seven at bats, and then instantly get the answer back. It comes back in tables and graphs, and then also human prose, narrative language.”
Developments in natural-language processing are beginning to allow for dialogue between people and A.I., which in turn creates a foundation for learning. Many people point to the bright, shiny object of vehicle automation when thinking about the potential for A.I., but perhaps the most exciting aspect of A.I. overall is the future of conversation and the amazing opportunity for learning quality exchanges between people and computers will provide. After all, this thirst for learning, and our need to talk to do so, may draw humans and computers even closer together.
On this episode of IT Visionaries, Corey covers the bases about how natural-language processing is being incorporated into the sports world, with professional leagues such as the NHL and beloved publications like Baseball America relying on the technology to get information to audiences more accurately and quickly. And, as Corey says, that’s just the beginning for Pramana Labs as its applications are seeping into other industries spanning from commerce to finance to mortgage lending. Enjoy the episode!
Takeaways
- Building a MVP: When creating a product, it’s important to solve one particular problem for a customer rather than trying to solve all of them. Lean on the customer to inform the scope of the product based on what they need.
- Making the NLP Reusable: Having reusable, paramentixed, and interchangeable pieces of NLP data cuts down the time required to get a question answered. Once a question is answered, another question can be asked with only a slight variation, and then another answer can be quickly provided. In this sense, both speed and delivery of accurate information increases.
- Answering Questions for Intent: The key in NLP is to discover what question a person is trying to ask and then offer up the correct answer. Looking at the entire sentence signature through an analysis of all the pieces that the computer has been trained for allows A.I. the ability to decipher the sentence and then respond accordingly.
- A True Two-Way Conversation: The future of NLP is really tech that scans a database and then provides prose, narrative responses. In fact, A.I. will react to things that are happening in real time and create a narrative for what it is seeing without the user necessarily having to ask any questions. This will allow A.I. to initiate conversation and guide a person to the knowledge that they are seeking.
For a more in-depth look at this episode, check out the article below.
Article
We talk to our computer by asking it a question and then the computer talks back, trying its best to answer it. At least that’s how the conversation generally happens between humans and A.I. Sometimes we’re happy with the computer’s answer and other times, not so much, and we either try again until we’re successful or give up. But real learning takes place with true dialogue, when there are successive exchanges that deepen understanding and where either a person or a computer can start the conversation. For people and computers to learn, good data is very important — as are ways to access it quickly. Even more important is constructive communication driven by language. Corey Patton is the Co-Founder and CEO of Pramana Labs, and he thinks prose and narrative conversations are the future in communication between A.I. and people.
“We created a way to learn about a relational database using training processes and NLP models that allows a user to just ask the question in free text,” Patton said. “What are the most home runs that any Angels outfielder has had in seven at bats, and then instantly get the answer back. It comes back in tables and graphs, and then also human prose, narrative language.”
Developments in natural-language processing are beginning to allow for dialogue between people and A.I., which in turn creates a foundation for learning. Many people point to the bright, shiny object of vehicle automation when thinking about the potential for A.I., but perhaps the most exciting aspect of A.I. overall is the future of conversation and the amazing opportunity for learning quality exchanges between people and computers will provide. After all, this thirst for learning, and our need to talk to do so, may draw humans and computers even closer together.
On a recent episode of IT Visionaries, Patton covered the bases about how natural-language processing is being incorporated into the sports world, with professional leagues such as the NHL and beloved publications like Baseball America relying on the technology to get information to audiences more accurately and quickly. And, as Patton said, that’s just the beginning for Pramana Labs as its applications are seeping into other industries spanning from commerce to finance to mortgage lending.
“Trying to figure out features versus customer versus customization versus applicability across multiple different leagues and even multiple different verticals is really where you want to look at when you’re building a product,” Patton said.
Patton then explained how a product must be built that has enough features to solve one minor problem, as opposed to trying to solve for a host of issues.
“You say, ‘Okay, I need an MVP,’” Patton said. “‘I’m attacking the sports space. I’m going to start with a league. I need to get a league that sees the need for this. I need to then figure out how I can create enough features so that it solves one minor problem.’ Not all of them. You don’t want to solve every problem because that’s not why you’re being engaged with from a league. You need to work with the customer. It all starts with them. It starts with engaging with a customer to say, okay, here is a pain point of pinpoint accuracy that you can then solve. And from there, every feature rolls out from engaging with them.”
In terms of Pramana’s work in NLP, Patton described two primary areas. The first involved making language reusable for the A.I. in a database. This effort greatly increases speed. As an example, he explained this in terms of Pramana’s work involving the NHL’s database.
“We made [the NLP data]it reusable in that when you started asking a question, it wasn’t just the most goals with a hard number that one team had scored in a specific period over a specific number of game series,” Patton said. “All of those pieces were all parameterized and interchangeable. You could ask one question and our training models knew that this entity was a goal. This was a period. This was a player. This was a team. And those were all interchangeable for asking one question forever and that never had to be written again. It just cut down the time for sending a need for an answer, to getting an answer […] and then changing something in that question to apply to another team, player, venue. And now you have another instant answer.”
The second main area in Pramana’s work includes A.I. providing prose, narrative responses. The conversation can be initiated by humans, but the technology enables the A.I. to start the conversation by analyzing its data and curating information such as potentational in-game storylines that it believes the users want to hear.
“It’s a two-way conversation,” Patton said. “We allow the user to ask a question in free text but we also have a full end-to-end pipeline with NLG technology that scans your database and pushes out prose narratives. It just sees things that are happening in your database and creates a prose narrative for them.”
The future of natural-language processing is very exciting. If A.I. has the ability to process information in real time and then can start conversations with humans, the possibilities are endless. This means that true conversation can develop between computers and people that only deepens over time. With two-sided prose, narrative conversations, learning and interdependence will only increase between people and computers.
To hear more about how Patton and Pramana Labs are supporting instructive conversations between people and A.I. check out the full episode of IT Visionaries!
To hear the entire discussion, tune into IT Visionaries here.
