With NLU technology provided by nuance.com, a software we got acknowledged with in the WearHack event, a idea of using speech to text software to make a virtual staff came up to our mind.
What it does
It simulates as a staff in Subway that can take your order by talking to you and finally build a pizza based on your description,by doing this, imagine a scenario you are in rush and has no time to line-up for ordering food, you may use such a voice-command app to order your sandwich/hamburger in advance to go to the store, this will save the your time on buying your meal. More on that, instead of other conventional on-line food ordering App which you have to browser and watch your App all the time, you may voice-command to finish your order, this function may apply into the case you want to order something when you are driving, and we think our app can apply to all other fast food restaurant (MacDonald, Burger king) as well.
How I built it
Using java to build it's logic and database, and you Bolt to train build up and train the language understanding intelligent. And then combine and test them on our android device.
Challenges I ran into
It wasn't easy to get familiar with the Bolt. The training process and database design was really time consumed. To make the virtual staff seemed more natural, a lot of logic was used to realize its functionality. And all language bases cost a lot of time of us.
Accomplishments that I'm proud of
But, it really worth it. As the robot can talk to us and serve us like a real human, recognizing or voice and understanding the sentences, we are proud of the work we programmed.
What I learned
We learned speech to text and text to speech, and logic constructions in this project, including a lot experience of android application developing, uses of API etc.
What's next for Subway_Samaritan
To make it more intelligent, and can cover more cases, even learn scenarios on its own by feeding it with big data of real sentences. That would perfection its performance.