The field of natural language processing is gradually evolving, and one startup is channeling its ability to scan unstructured information about companies and business relationships and extract meaningful insight.

This can be especially helpful to sales teams, as Ogmagod Inc. is discovering in the rollout of its data-processing solution.

“What we do is help salespeople understand the customer better so they have a higher probability of winning the deal or shortening the sales cycle,” said Ben Cheung (pictured), co-founder of Ogmagod. “Salespeople are really put up with an impossible task because of all this information out there they are expected to know. We can use developmental machine-learning models and use what’s out there on the web. That’s what differentiates us.”

Cheung spoke with Jeff Frick, host of theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how the company uses machine learning to help sales teams, advances in natural language processing and the work that remains to move the field forward.

Generating potential opportunities

The name for Cheung’s startup comes from Ogma, the Celtic god of wisdom, speech and language. The company uses artificial intelligence and machine learning to combine all three areas of human capability and help salespeople in business-to-business marketing sift through a mountain of information.

“What salespeople look for is a quite narrow domain,” Cheung explained. “They are looking for problem statements within customers that relate to what they are selling. We’ll put them into our machine-learning pipeline and filter out all the junk. The output is a set of potential opportunities.”

To arrive at those opportunities, Ogmagod uses natural language processing algorithms to focus on information of specific interest to a client. The firm takes advantage of advances in machine learning that have moved natural language understanding forward.

“It’s a progression, but machine learning is a very powerful and popular technique right now,” Cheung said. “There have been a lot of breakthroughs in the last few years. With speech detection and image processing, there has been tremendous progress.”

Despite these advances, the field of natural language processing remains very much a work in process. Machines have not yet become adept at reading emotion and context, which form so much of human communication. The area where development has been slower involves natural language generation, according to Cheung.

“It means the machine knows how to compose sentences and generate them back to you. That is still in very early days,” Cheung said. “People say we’re at the cusp of being able to understand language in general. I don’t believe it; we’re very far away from that.”

There is also the problem confronted by startups in the field, which involves limits on the amount of available information to train machine-learning models. Not every company has access to information that a firm such as Google LLC can tap, and this led Ogmagod to focus on specific use cases using precise domains.

“You cannot expect to have the kind of dataset that Google or Microsoft or Facebook has,” Cheung said. “You have to deliver something with a smaller dataset, so you have to narrow your domain.”

Prior to co-founding Ogmagod, Cheung spent the past six years working on natural language processing solutions for Microsoft Corp. and Genee Inc. The goal is to help salespeople formulate a strategy that will win a deal by engaging with the customer at a level that sounds like they are employees of the prospect company itself.

“We’re taking away the time-consuming process of data collection and cleaning up the data,” Cheung said. “The real human intelligence should be focused on data analysis to derive lots of insights.”

Here’s the complete video interview, one of many CUBE Conversations from SiliconANGLE and theCUBE:

Photo: SiliconANGLE