Speech Technology Magazine


Api.ai Partners with Slack to Add Natural Language Understanding to Chatbots

Api.ai for Slack lets developers create conversational chat bots that can understand user intent.
By Tye Pemberton - Posted Mar 21, 2016
Page1 of 1
Bookmark and Share

Conversational user interface startup Api.ai (Speaktoit) has just announced a strategic association with Slack, a cloud-based collaboration startup. Api.ai's new platform, Api.ai for Slack, empowers Slack developers to create chat bots within the collaboration suite that can communicate more naturally, accurately parse text, and execute a variety of tasks, all with less work on the back end.

Api.ai CEO Ilya Gelfenbeyn and Teresa Dodson, vice president of marketing, see the new partnership as an ideal way to further showcase the power of their product's capabilities. "There are more users on messengers than on social networks," Gelfenbeyn said, noting that he sees Slack as a new platform for advanced natural language understanding.

Dodson expects the company to see a migration toward the use of Api.ai for enterprise applications.

Slack, launched in 2013, was originally a house-grown messenger tool for the company Tiny Speck, which was then working on the computer game Glitch. Since refocussing on the collaboration tool and renaming the company, Slack has grown to be used by more than 30,000 teams and was valued at $2.76 billion during its most recent funding round.

Speaktoit launched Api.ai, a natural language understanding platform that combines speech recognition, natural language understanding (NLU), and text-to-speech technologies, in September 2014.

"Two or three years ago chat wasn't that popular," says Gelfenbeyn. "Originally at Api.ai we were focused on voice control as a use case, but about a year ago we found that some of our clients were building Slack bots and tools for other messengers and realized what a great environment chat could be to show where Api.ai really shines. It helps us show prospective clients how good our natural language understanding technology is. There no issue with speech recognition accuracy to get in the way.

"At the time there were no special tools for developers who wanted to use Api.ai to build chat bots. Just seeing the growth, we decided it would be much nicer to have some specialization to simplify integration," he continues.

Since its inception, Api.ai has enabled developers to define parameters and context for responses from only a few examples of possible end-user inputs. These user intents are then corresponded against actions or responses the developer builds, after which Api.ai uses its machine learning and database of more than 2.6 billion intents to create a fully conversational agent. With the new platform, developers can then directly export the agent to Slack and create a chat bot with Api.ai's fluent natural language understanding in one click.

"Api.ai is very good at conversation management," says Gelfenbeyn. "End users can switch between topics, stop requests, and the platform is very good at figuring things out and asking for any information it still needs to give a developer-defined response to the intent."

Slack has also recently courted other platforms like Google Cloud, Microsoft Azure, and Amazon Web Services Lambda, which will be offering the collaboration service streamlined integrations as well, promising to increase the application's utility even further. While Gelfenbeyn recognizes that enterprising developers could cobble together a fully functioning intelligent chat assistant from the various Slack enhancements in the pipe when coupled with Api.ai, he does not feel like they threaten Assistant, Api.ai's best-of-breed conversational assistant application.

"Ideally," he says, "those interfaces can be universal. There's no reason to create an ultimate service. We see all of Slack's new integrations as potential partners for us."

Page1 of 1