Intent recognition is the natural language understanding (NLU) task of determining what general goal a user is trying to accomplish (e.g., finding out the weather forecast, booking a table at a restaurant, or adding a song to a playlist). What's tricky is there are many ways users may express an intent. For example, "Turn on the light" and "It's too dark in here; make it brighter" are just two of a plethora of ways of expressing the same "Light on" intent to a smart home device, but the two utterances are completely different on the surface in terms of syntax and vocabulary. A good intent recognizer should map both of those utterances to the same intent. More generally, a well-trained recognizer can account for the many ways people may express their goals in natural language and map them to the correct intent, which then triggers an action or response.

With Enterprise Connect 2018 fast approaching, you're no doubt doing a lot of planning to prioritize which meetings to schedule and which sessions to attend. You can't do it all, and this is my moment to draw your attention to the Speech Technology track, a new addition to the EC lineup. In this inaugural year, the Speech Tech track may not yet be on your radar. I'm hoping this post will change that, especially since I'm kicking off the program with a tutorial on enterprise speech technology on Monday, March 12, at 8:00 a.m. If you like what I have to say, you'll probably want to attend more sessions for this track, and that will help validate the move to put speech tech on the program.

Facebook initially focused on ads. The rationale was that at the time people were typically scrolling through their feeds with the sound off, so for advertisers to get their message across they needed text to run inside their video ads. "We looked for a problem space in the speech recognition area through which we could deliver value to users," said Reena Philip, an engineering manager for Facebook's speech infrastructure group. Joining forces with the ads team, "we collaborated closely on a prototype," she said. The feature launched in the second quarter of 2016.