What it does
This tool allows customer support systems to automatically create responses based on messages sent to the Bell twitter account. It will take in the message, classify it using Natural Language Processing techniques, and then respond to the user. It will make it easier to respond to tweets, and direct customers in a more efficient manner.
How we built it
We used TensorFlow, Scikit-learn, NLTK, and Pandas to implement a real-time request categorization and customer support system based on customer tweets.
Challenges we ran into
Finding enough data to adequately train a model was tough. We used data augmentation to create some phrases from what we were given. Training on the time limitations was also tough as we wanted to train with an LSTM, but didn't have time or enough data.
Accomplishments that we're proud of
We made a working model! It will take in a user's name and tweet and automatically generate some output. The accuracy was >85% for train and test too.
What we learned
We learned how to use libraries such as Scikit-Learn and NLTK to implement our idea.