search

enzoampil / tito-joker

A humorous AI that uses state-of-the-art deep learning to tell jokes
http://35.225.94.177:8501/
GNU General Public License v3.0
45 stars 5 forks source link

[experiment] Apply generation techniques employed from “abstractive summarization” and “Answer generation from Q&A augmentation” #18

Open enzoampil opened 4 years ago

enzoampil commented 4 years ago

Foundational source for abstractive generation: https://arxiv.org/abs/1704.04368

enzoampil commented 4 years ago

Lessons to draw from controllable image generation:

https://blog.insightdatascience.com/generating-custom-photo-realistic-faces-using-ai-d170b1b59255

enzoampil commented 4 years ago

Respecify joke generation as a seq-to-seq generation model

enzoampil commented 4 years ago

Encoder decoder can be done w/ summarization specification

enzoampil commented 4 years ago

Leaning towards a T5 specification. Can write a new blog post about this with title "Conditional Joke Generation with T5".

https://arxiv.org/pdf/1910.10683.pdf

enzoampil commented 4 years ago

Related issue about paraphrase generation:

https://github.com/huggingface/transformers/issues/3725

enzoampil commented 4 years ago

T5 doc (refer to SQuAD specification in Appendix D)

https://huggingface.co/transformers/model_doc/t5.html

enzoampil commented 4 years ago

https://arxiv.org/pdf/1911.00536.pdf