NLP News Cypher | 04.26.20

Author(s): Ricky Costa

Originally published on Towards AI.

NATURAL LANGUAGE PROCESSING (NLP) WEEKLY NEWSLETTER

NLP News Cypher U+007C 04.26.20

Recursion

Last week was a whirlwind. Thank you for your support. U+1F92F

If you’re not in the loop, last week we released the Super Duper NLP Repo! A collection of 100+ Colab notebooks using models to perform various NLP tasks. Periodically, we’ll continue to update this repo as we've done with the Big Bad NLP Database. If interested, you can always follow our Twitter for database updates.

To explore the repo, travel here:

Oh and good news, Colab’s project manager likes the Super Duper NLP Repo:

FYI, we’re prepping for our upcoming update of the Big Bad NLP Database, and if you know of a dataset to include, please let us know! We’ll give you a nice shout-out on Twitter.

How was your week? U+1F60E

This Week:

AI Goes Virtual

Serving with Facebook & AWS

ScaledML Conference Presentations

DeepPavlov Updates

Paraphrasing Conversational AI

Dataset of the Week: SCAN

AI Goes Virtual

As you read this, the ICLR conference is currently underway (virtually). In the meantime, you can always check out their accepted papers in the link below. Paperswithcode released a nice list of ICLR papers, which of course, includes code!

List of ICLR Papers:

Most Popular Paper:

Serving with Facebook & AWS

Looks like PyTorch is getting a partner with the folks that deliver us those brown boxes every day. With more and more peeps looking to deploy models with their favorite framework, AWS and FB delivered TorchServe this past week (and it’s implemented in Java):

TorchServe is a framework for serving AI models and offers the developer with…

“multi-model serving, logging, metrics for monitoring, and the creation of RESTful endpoints for application integration.”

Example codebase:

GitHub:

Google be like…

ScaledML Conference Presentations

Back in February, when traveling was still a thing, a conference happened. And all the stars showed up: Chollet, Karpathy, Sutskever and many others.

(Even Lex Fridman’s Men-in-Black suit showed up)

In the link below, you can find their presentations and if you’re lucky, a few of them provide PowerPoint slides.

My personal favorite prese comes from Karpathy because we get a first-hand account of real-world ML models at scale. And according to the Tesla guru, just troubleshooting stop signs is a major headache.

DeepPavlov Updates

Don’t sleep on DeepPavlov, they’ve got the goods when it comes to NLP. So much, that I’ve added their great notebooks to the Super Duper repo.

And they also have a new update for their framework:

• A dataset reader for the IMDB Large Movie Review Dataset and sample configurations for training classifiers on it
• Models for speech recognition and synthesis based on NVIDIA NeMo
• A New hybrid NER model pre-trained for English and Vietnamese languages
• A pre-trained NER-based Model for Sentence Boundary Detection Task

Paraphrasing Conversational AI

RASA is experimenting with a paraphrasing model in order to improve the intent classification for conversational agents. Their model takes in a sentence and then generates multiple paraphrases. If you use it, make sure you leave RASA a note as they’re looking for community feedback. (they’ve included a Colab notebook too)

Colab of the Week:

Dataset of the Week: SCAN

What is it?

“SCAN is a set of simple language-driven navigation tasks for studying compositional learning and zero-shot generalization.”

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