Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/enzoampil/tito-joker
A humorous AI that uses state-of-the-art deep learning to tell jokes
https://github.com/enzoampil/tito-joker
Last synced: 2 days ago
JSON representation
A humorous AI that uses state-of-the-art deep learning to tell jokes
- Host: GitHub
- URL: https://github.com/enzoampil/tito-joker
- Owner: enzoampil
- License: gpl-3.0
- Created: 2019-09-23T01:58:15.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2023-06-12T21:27:20.000Z (over 1 year ago)
- Last Synced: 2024-11-01T04:51:54.459Z (9 days ago)
- Language: Jupyter Notebook
- Homepage: http://35.225.94.177:8501/
- Size: 6 MB
- Stars: 45
- Watchers: 3
- Forks: 5
- Open Issues: 26
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Roadmap: ROADMAP.md
Awesome Lists containing this project
README
# Hi, I am [Tito Joker](http://35.225.94.177:8501/)! :wave: :grinning:
[](https://www.gnu.org/licenses/gpl-3.0)
[](https://travis-ci.com/enzoampil/tito-joker)
[](http://35.225.94.177:8501/)
[](https://github.com/ambv/black)
## A humorous AI that uses state-of-the-art deep learning to tell jokes
**Tito Joker** aims to understand humor well enough to tell jokes that are actually funny. All you have to do is input a riddle type question and he tells a joke using it. He still has a long way to go but we will get there!
Interact with Tito Joker on this [website](http://35.225.94.177:8501/) and check out his [medium article](https://towardsdatascience.com/can-a-robot-make-you-laugh-teaching-an-ai-to-tell-jokes-815f1e1e689c?source=friends_link&sk=342342be4cbf0064f8f0e7cb7ec0b6bc)!
## Generate GIFs automatically
*Turned off by default - turn on from left sidebar*
## Tell multiple jokes from a single input
*Tito Joker can give unique responses to the same input*
## Joke rating system
*Help Tito Joker improve his humor by giving him feedback!*
## Methodology
### Architecture
Fine-tuned version of the recently released [OpenAI GPT-2 model](https://openai.com/blog/gpt-2-1-5b-release/) with a left-to-right language modeling training objective. Similar hyperparameters were used from the [original paper](https://d4mucfpksywv.cloudfront.net/better-language-models/language_models_are_unsupervised_multitask_learners.pdf).
### Data
A [jokes dataset](https://www.kaggle.com/abhinavmoudgil95/short-jokes) from Kaggle was used for fine-tuning. Aside from the original preprocessing, additional special tokens were added to allow the model to understand the difference between the "question" and "answer" components of a riddle type joke.
*Warning: the dataset contains NSFW jokes, so Tito Joker's humour will also reflect jokes of this nature.*
## Preparing the data
```
python ./src/utils/process_jokes.py shortjokes.csv ./data/riddle_jokes.txt
```
## Training the model
Please refer to the experiment [notebooks](https://github.com/enzoampil/tito-joker/tree/master/experiments).
*Note: The number at the end of the notebook's filename corresponds to the model version it produces (no number means 1)*
## Trained models
1. [Tito Joker v1 (OpenAI GPT-2)](https://storage.googleapis.com/joke-generator-model1/model1.zip)
2. [Tito Joker v2 (OpenAI GPT-2)](https://storage.googleapis.com/joke-generator-model1/model2.zip)
## Acknowledgments
Special thanks to [Hugging Face](https://huggingface.co/) for their implementation of OpenAI GPT-2 using PyTorch, [Streamlit](https://streamlit.io/) for making it super easy to deploy Tito Joker as a webapp.
## About
**Why is the AI named Tito Joker?** Because in Filipino, "tito" means "uncle" when translated to English, and in the Philippines, we all have that uncle who says the corniest jokes!
To learn more, please refer to Tito Joker's [medium article](https://towardsdatascience.com/can-a-robot-make-you-laugh-teaching-an-ai-to-tell-jokes-815f1e1e689c?source=friends_link&sk=342342be4cbf0064f8f0e7cb7ec0b6bc) on Towards Data Science.