Alibaba Machine Intelligence Technologies today unveiled its latest intelligent speech interaction technology aimed at “smart ordering” in cafés and restaurants.
The Alibaba DAMO Academy unit, which focuses on fundamental AI research, said the new technology lets buyers, for example, order their coffee by speaking to a smart ordering machine. The machine will understand their requests and display the order onscreen in a split second. Buyers can correct their orders as many times as they want, and the machine is expected to catch all the changes and update immediately.
For example, a customer might say: “I want to order two large cups of cappuccino. Oh, please make them with less sugar and decaf. To go. And sorry. That should be three cups of cappuccino, two large and one small.” The smart ordering machine will then display the full order as two large cups and one small cup of decaf cappuccino, all with less sugar than usual and note that it’s for takeout.
Smart ordering is possible through the team’s latest research in spoken language understanding (SLU), a field that involves both speech processing and natural language processing. First, the team develops a multi-modal speech interaction solution that can capture both voice and visual features, taking into account such things as the speaker’s pace, pauses between words, pronunciation, breaths and facial expression. Next, the team builds a reinforcement-learning model that allows for revisions and intent detection. All are important in the SLU field, making it possible to apply the latest SLU technology into commercial use.