We developed several chabots using the pretrained model of DialoGPT from transformer library of 🤗 Hugginface by performing fine-tuning of its small version on some corpus of data coming from tv show and movies, for different characters.
We choose to extend the work made by [Nguyen et al.] (https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1174/reports/2761115.pdf) who explored the task by implementing a chatbot by a seq-to-seq model. This work is by all means a revision and extension of theirs.
| Character | TV show/movie |
|---|---|
| Barney Stinson | How I Met Your Mother |
| Sheldon Cooper | The Big Bang Theory |
| Joey | Friends |
| Phoeby | Friends |
| Harry Potter | Harry Potter |
| Fry | Futurama |
| Bender | Futurama |
| Darth Vader | Star Wars |
Please install all dependencies within "requirements.txt" through pip. There is also a GPU version for these same requirements, but it may is CUDA dependant.
The list of relevant folders for this repository is:
Datafolder contains all the data we used to fine-tune our models and where we saved the models,Codefolder contains the notebooks and also the custom libraries useful to compute metrics and plotting,Metricsfolder which contains metric results in json format and plots
As an extension of the original project, we implemented and tested a variety of metrics to evaluate the performances of these chatbots.
We also deployed a drive folder which already contains the trained models, you can check it following this drive link