Generating lyrics using recurrent neural networks (with David Volquartz Lebech)
Details
We are very excited to have David Lebech, Senior Data Scientist @ Goodiebox present a talk on natural language generation!
"In recent years, models such as Open AI's GPT and Google's BERT have taken the spotlight when it comes to generating text with deep learning. The success of the Transformer-based architecture and larger and larger models have happened so quickly and shown such convincing results that even the mainstream media has taken notice. But while transformer architectures seem to have become the de facto standard for language models across many domains, recurrent neural networks (RNNs) are a less recent invention that still work very well for many use cases.
In this talk, we take a brief look at how a recurrent neural network works and use it for generating lyrics with a model architecture that is possible to train on a normal CPU with a bit of patience."
You can check out David's excellent blog along with his open-source projects at https://davidlebech.com
We will be recording the session's presentation section due to multiple requests, but we will not be recording the subsequent discussion at the end, so please make sure you join the live event, if you can, not to miss out!
All resources and presentation recordings are shared via our slack group. Feel free to stop by and say hello: https://join.slack.com/t/nlp-studygroup/shared_invite/zt-o29hsx83-BA2MDvShaCYVF8yW2sGvdQ