Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/mdh266/jfkspeechwriter
https://github.com/mdh266/jfkspeechwriter
data-science google-cloud keras machine-learning natural-language-processing nlp recurrent-neural-networks tensorflow2
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/mdh266/jfkspeechwriter
- Owner: mdh266
- License: mit
- Created: 2022-12-05T00:54:51.000Z (about 2 years ago)
- Default Branch: main
- Last Pushed: 2024-01-31T02:35:43.000Z (12 months ago)
- Last Synced: 2024-06-11T17:54:00.582Z (7 months ago)
- Topics: data-science, google-cloud, keras, machine-learning, natural-language-processing, nlp, recurrent-neural-networks, tensorflow2
- Language: Jupyter Notebook
- Homepage:
- Size: 5.79 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Creating An AI Powered JFK Speech Writer
-----------------
## Part 1
In this short post I went over how to scrape the President John F. Kennedy Library's website to create a collection of JFK speeches. I covered how to do this using [BeautifulSoup](https://www.crummy.com/software/BeautifulSoup/bs4/doc/) and upload them as text files to [Google Cloud Storage](https://cloud.google.com/storage). One thing I could have done is to use an asynchronous HTTP client [AIOHTTP](https://docs.aiohttp.org/en/stable/) to read and write using asynchronous I/O.
## Part 2
In this blog post I covered how to create a generative text model using bi-directional [gated recurrent unit (GRU)](https://en.wikipedia.org/wiki/Gated_recurrent_unit) that is trained on speeches made by President John F. Kennedy. The model was built in [Keras](https://keras.io/) using [TensorFlow](https://www.tensorflow.org/) as a back-end and I covered how to use this model to generate text based off an input string.
The GRU model is a specific type of [Recurrent Neural Network (RNN)](https://en.wikipedia.org/wiki/Recurrent_neural_network) and models sequences. RNNs were quite popular for Natural Language Processing until around 2017/2018. More recently, Recurrent Neural Networks have fallen out of popularity for NLP tasks as Transformer and Attention based methods have shown substantially better performance. Using transformers for generating text that is meant to sound like JFK would be a natural next step and will be a follow up for a future blog post!