Hi! I am happy to announce that the NLP Newsletter is back! This iteration emphasizes on AI research, perspectives, technologies, and education.
Welcome to the 33rd Issue of the NLP Newsletter! Here is a recap of the recent notable NLP and AI news: AI talent supply vs. demand, radiological analysis, biased face recognition, several deep learning educational material, conversational AI, and much more.
On People…
The 2019 Global AI Talent report has been released with the main finding being that supply for top-tier AI talent does not meet the high demand — Link
Renowned AI researchers are asking Amazon to stop selling its facial-recognition to government agencies due to well-known bias issues — Link
A machine learning paper proposes an approach, inspired by psychological observations, which uses comparisons to conduct inference on unlabeled data and feed this information into the learning model. It is inspired by human learning and aims to reduce the need for human-annotated data — Link
Can machine learning systems identify biomarkers for lung cancer survival? Read more on the topic in this latest work on tumor segmentation— Link
On Research….
Learn how NLP models are vulnerable to paraphrase attacks. Rada Mihalcea commented on Twitter that “we need to start thinking about how to build systems that are not only accurate and efficient, but also robust to potentially malicious inputs” — Link
Sebastian Ruder, now at DeepMind, publishes his dissertation on “Neural Transfer Learning for NLP” — Link
This paper reviews the recent advances in machine learning and its applications to radiological images. It covers trends, perspectives, and topics such as medical image segmentation and neurological disease diagnosis — Link
Distill published a new online paper called “A Visual Exploration of Gaussian Processes” that aims to visually demonstrate and guide researchers on how to turn a collection of building blocks to solve regression problems — Link
This paper provides a comprehensive overview of machine learning and NLP models and their usage in the healthcare space — Link
On Education…
AI instructor, Alexis Cook, releases a set of Kaggle notebooks (structured as a course) on learning how to program and make great data visualizations using Python. The course is free and consists of 4 hrs of material and 15 lessons — Link
“Dive into Deep Learning” is an amazing resource for learning machine learning and deep learning concepts via an online interactive book — Link
Pieter Abbeel will be teaching a new course on “Deep Unsupervised Learning” with topics ranging from ImprovedGAN to PixelCNN++, and much more. Video lectures are publicly available — Link
Stanford CS231n 2019 updated course (Convolutional Neural Network for Visual Recognition) syllabus is now available online — Link
Stanford also releases course material for CS224n (NLP with Deep Learning); includes video lectures, slides, and notes — Link
This article discusses how Swisscom uses TensorFlow to conduct real-time scalable chat intent and email classification — Link
On Code and Data…
DeepMind releases a dataset containing mathematical question and answer pairs intended to test the mathematical learning and algebraic reasoning of neural network models — Link
PyTorch BigGraph is a distributed system for learning and generating large graph embeddings from large-scale graph-structured data — Link
Trung Tran shows us, in this detailed blog post, how to train neural machine translation models with attention using TensorFlow 2.0 — Link
On Industry…
Microsoft launches a free online AI Business school. The course focuses on teaching about AI strategies and culture, designed to improve AI leadership in technology companies— Link
HuggingFace research scientist, Victor SANH, discusses how they use state-of-the-art NLP algorithms for building conversational AI systems — Link
Worth Mentions…
Mathpix is a tool that allows you to instantly convert images into latex — Link
You can now launch Colab notebooks directly from PyTorch’s tutorials — Link
Mathigon is a fun and interactive website that aims to teach mathematical concepts using a visual and interactive approach — Link
Serena Yeung chats with MIT Technology Review about her experiences of building AI-based healthcare systems for the real world — Link
Jay Alammar writes an impressive blog post discussing the inner works of word2vec; topics include language modeling, skip-gram, and others — Link
Distill publishes an article on how to visualize memorization in RNNs — Link
