This is a chatbot created with deep learning modeled off of the TV character Squidward from Spongebob. The model was created using Python and PyTorch and the model was deployed using a Flask application on Heroku.
The first model was not very coherent so I added in the lines from the other characters. The first model was trained on my computer using CPU only. I also trained a couple more with slightly different parameters using an AWS EC2 instance.
- Created a script to extract the transcript for every episode from website into a .txt file [generateTranscripts.py]
- Created functions to extract sentence-response pairs for a specified character [parse_transcript.py]
- Defined classes for holding and proccesing vocab and sentences [load_data.py]
- Defined model classes and ran data through models [model.py]
- Integrated model into Flask application [app.py]
- Optimize models for better performance
- Fix inconsistent punctuation
- Improve Frontend design
- The model is too large to upload to GitHub so figure out workaround