OskiChat

Inspiration

Tired of long lines waiting for your scholarship to be processed? Look no further, OskiChat is here to bear your pain. We were inspired by our own frustrating experiences with advising and the many questions which would have had straightforward answers if all the information was centralized.

What it does

OskiChat is here to answer every Cal students' burning advising questions, from when your next semester class is held to information about your in-process transcript.

How we built it

We used python NLP and ML libraries to train a model to respond to a set of questions. OskiChat also incorporates web scraping of classes.berkeley.edu to provide students with up-to-date information on the classes being offered next semester. We used the Flask web-app, with js and jquery, to integrate this into a dynamic website where students can interact with OskiChat.

Challenges we ran into

We really struggled connecting the back-end of the website through Flask with the front-end javascript designs from scratch since we decided against using existing chatbot APIs. It took a lot of bug-hunting and discovering that documentation was outdated before we figured out how to successfully send HTTP REST requests.

What we learned

We learned how to connect the front end to the back end, the basics of training a machine learning model on text, and web design from scratch. CalHacks 6.0 also taught us how to flesh out a product development timeline and work as a team.

What's next for OskiChat

We would really like to flesh out our current design with more features! These will include follow-up questions, a more robust machine learning database, transfer credit questions, as well as determining if a student's four year plan meets all their graduation requirements.

Built With

• bs4
• css3
• flask
• google-app-engine
• html5
• javascript
• jquery
• nltk
• numpy
• python
• tensorflow
• tflearn