AREL, TextWorld, eCommerce Chatbot, Bloomberg and BAIR AI Courses, Troubling Trends in ML, ICML 2018 Notes, NLP and Law,…
Welcome to the 21st Issue of the NLP Newsletter! Here is this week’s notable NLP news! The best of ICML and ACL, new AI courses, lots of guides on things like evaluating feature importance and feasible human evaluation, story evaluation, reproducing deep learning papers, and much more.
On People…
A new paper presented in ICML 2018 describes some of the troubling trends in machine learning (ML) scholarship. The authors, Zachary C. Lipton and Jacob Steinhardt, discuss the implications of these trends on the field of ML and the proper distilling of machine learning research and ideas — Link
“NLP is allowing humans to interact with computers the same way they interact with themselves. It is also increasing the use of computers in our lives. Hence man-machine interface is not only changing, it’s importance is actually reducing” — Link
In machine learning we rely a lot on human evaluation but how do we make it cheaper, here is a nice article explaining how this can be done — Link
Excellent notes on ICML 2018 presentations — Link
On Education and Research…
Find out the math insights behind 10 type of GANs (video) — Link
Berkley AI Research (BAIR) offers new “Intro to AI” course — Link
On evaluating interpretability methods that estimate feature importance in deep neural networks — Link
Bloomberg is offering new course on “Foundations of Machine Learning” — Link
Learn how to apply machine learning for music exploration and analysis — Link
AREL is an approach that uses adversarial reward learning for storytelling and addressing limitations of the metrics used for story evaluation (code + paper) — Link
The University of Texas if offering a new course on Linear Algebra hosted on edX — Link
On Code and Data…
Microsoft releases a massive collection of free datasets for advancing NLP and machine learning research — Link
On challenges related to the reproducibility of machine learning algorithms and research (slides) — Link
15 Tensorflow modules to generate images using GAN (notebook) — Link
On reproducing a deep learning paper and everything you need to know about this process — Link
On Industry…
Learn how law firms are using NLP to change the way they do business — Link
Facebook files patent for eCommerce messenger bot with NLP being the core technology behind the technology — Link
Samsung’s CoZNet, a deep reinforcement learning algorithm, achieved SoTA for two popular NLP challenges — Link
Learn how Microsoft is using inspirations from text-based games to create a learning environment for training reinforcement learning agents — Link
Quotes of the day…
Worthy Mentions…
Debunking batch normalization — Link
NLP Newsletter — Issue 20: Grokking Deep Learning, Something-Something V2, Tensorflow 1.9, NLP ImageNet Moment, Feature-Wise Transformations — Link
Machine learning roundup — Link
If you spot any errors or inaccuracies in this newsletter please comment below. You can also reach me on Twitter if you want to discuss any of the items above.