NLP News Cypher | 06.28.20

Author(s): Ricky Costa

Originally published on Towards AI.

NATURAL LANGUAGE PROCESSING (NLP) WEEKLY NEWSLETTER

NLP News Cypher U+007C 06.28.20

The Final Frontier

Geoff Hinton backpropagated himself and cancelled a bunch of meetings. He realized that his idea on perceptual learning had a central flaw and went scorched earth on his planned talks. Hinton doesn’t mess around.

Last week was a blur…

…oh, and there’s a new supercomputer in Japan:

The Big Bad NLP Database was updated U+1F60E. We added 63 new datasets, taking us to 545 total! Thank you to all contributors: Erik Chan, Ramya Tekumalla, Monisha Jegadeesan, and Purva Tendulkar!

P.S. there’s a U+1F680 button at the bottom of the page if you want achieve escape velocity.

Lastly, we updated our homepage, inspired by Darkwing Duck and Jean-Luc Picard. If you need NLP software, drop us a note! U+1F601

This Week:

Deep Learning Drizzle

Colab Tricks and Treats

Colab of the Week

Interactive Tabular Data

Another AI Survey

Dataset of the Week: ClarQ

Colab Tricks and Treats

In this industrious blog post, Amit Chaudhary shows us all the magic tricks that Colab has to offer. In total, there are 17 awesome hacks you can try out such as: Jupyter notebook keyboard shortcuts, running flask apps from Colab, managing Colab from command line (just to name a few!) and many more.

Colab of the Week

While we’re on the subject of Colab, Jason Phang built a great Colab for using the HF libraries for multi-task training. The project involves using 3 datasets for 3 separate tasks: semantic text similarity, natural language inference (NLI) and multiple choice QA. Jason highlights the creation of 3 different models while sharing the same encoder. And this is awesome because “the shared encoder ensures that during training, all updates will update the same encoder weights, and also does not consume any additional GPU memory.”

Deep Learning Drizzle

If you like to watch YouTube tutorials/ lectures there’s a major database that aggregates such videos by domain. Enjoy!

Interactive Tabular Data

While Wolfram is not searching for a Theory of Everything he’s gotta run a business. In a new Wolfram blog post, they explore new capabilities for manipulating Datasets (tabular dataframes) in Mathematica 12.1.

Another AI Survey

Several 100 C-Suite peeps were surveyed regarding their use of AI in the enterprise.

TL;DR

-CTO’s make all the decisions.

-Azure is #1 AI Cloud provider.

-Text is the most widely used data type.

-Biggest bottleneck in working with AI is ‘lack of resources/talent’

Dataset of the Week: ClarQ

What is it?

Dataset used for clarification question generation consisting of ∼2M examples distributed across 173 domains of StackExchange.

Sample:

Where is it?

Every Sunday we do a weekly round-up of NLP news and code drops from researchers around the world.

If you enjoyed this article, help us out and share with friends!

For complete coverage, follow our Twitter: @Quantum_Stat

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI