Building a Pipeline for State-of-the-Art Natural Language Processing Using Hugging Face Tools

The natural language processing (NLP) landscape has radically changed with the arrival of transformer networks in 2017. From BERT to XLNet, ALBERT and ELECTRA, huge neural networks now manage to obtain unprecedented scores on benchmarks for tasks like sequence classification, question answering and named entity recognition. The pipeline from text to prediction remains complex, but tools like huggingface/transformers and huggingface/tokenizers take most of the burden off of the user, offering a simple API. This talk will focus on the entire NLP pipeline, from text to tokens with huggingface/tokenizers and from tokens to predictions with huggingface/transformers.

About:
Databricks provides a unified data analytics platform, powered by Apache Spark™, that accelerates innovation by unifying data science, engineering and business.

Connect with us:

[EvanZamir]

Am I crazy or did this tutorial not explain how to use these models in Spark? Specifically, I'd like to use a model to perform inference on text that is in a dataframe.

[goldnutter]

Thankyou for letting me listen and.. my mind breaking a bit and reforming lol

I will have to listen to this one in a few years and look at where GPT^3 goes

[goldnutter]

API world

the uses of this.. for humanity.. to find edge cases in language, meaning.. expression

to bring the world closer together and understand language and reality itself

this is only the beginning

[goldnutter]

a vote doesn't break a 50:50 tie

but when you have triangles you can.. essentially have information now..

[goldnutter]

18:00 the holy grail, character tested randomness test in code

Not easy.. easy for a human

[trippygod]

*Databricks* my beats would sound great on your channel 💯🔥🔥🔥 lets collab

[goldnutter]

16:00 or so .. very large vocabulary unlike me

backtest @ nonsense speak