Convenience and engaging social experience: these are two factors which truck drivers often lack on the job. Given the considerable rise in the demand for these employees around the nation, we wanted to develop a chatbot which would help motivate them to take on this infamously tedious and isolating task.
How we built it
Google Cloud Speech-To-Text API, Spacy NLP software, and Abundant Training Data
Challenges we ran into
One of the most difficult aspects of this application was developing a solid training model with our data set, how could we make something which is both entertaining for the user and convenient when needed? So we made a choice on which elements of our chatbot would serve most beneficial to our target user.
Another challenge was figuring out how to make a connection to Google's Cloud Services in a way that would increase our chatbot's performance especially when it came to learning from a massive dataset. We wanted to run the entire bot from a GCP VM, as this would streamline efficiency and ease of use, however streaming input speech became a challenge when dealing with the cloud VM. Ultimately we determined a reliable _ local _ build would be most practical at the cost of computing efficiency.
Accomplishments that we're proud of
Succeeding Outside of Our Comfort Zone
- We didn't really know how we could implement machine learning or NLP given our previous experience, it was totally new and felt amazing to produce something that was doing what we wanted
What we learned
A Functional Product Will Always Have Potential
- We know our product isn't perfect but accomplished plenty and are aware of the room for growth this product provides
Know When to Ask for Help
- Machine Learning is a difficult topic to work with, it's important to acknowledge when you are stuck as many times it can be the only path towards progress and completion of our goals
What's next for RicottasTransport
- A way to provide check-in and local stops where other truck drivers are staying, attending to the social needs of these workers and establishing community among them
- We look forward to implementing ways for our chatbot, Ricotta, to provide suggestions for these travelers in real time for places they can visit for a quick bite or routes/detours they can take given the ever-present constraint for time