decaNLP, GluonNLP, al+ stack, Pythia, Data for Good, Differentiable Image Parameterizations,…
Welcome to the 23rd Issue of the NLP Newsletter! Here is this week’s notable NLP news! Lots of NLP tools, a couple of surveys discussing the current state of AI and NLP, an introduction to transfer learning for NLP, Tensorflow and PyTorch guides, and much more. Enjoy!
On Society…
Recent survey produced interesting results that reflected the different views of people on AI technologies. For instance, 61% of people agreed that educating people about AI technologies is a responsibility that should be shared — Link
Learn about how NLP can be used to detect cyber attacks through understanding deep thoughts and linguistic cues found in resources already available on the internet (e.g., hacker chatrooms) — Link
Rachel Thomas gives an overview on strategies to analyze and prevent bias in machine learning research — Link
Sara Hooker discusses the issue of “data for good” — Link
Jacob Andreas shares his perspective on meaning representations in models without logical forms — Link
On Education and Research…
The Natural Language Decathlon (decaNLP) benchmark offers a new and unique setting for studying general NLP models that can perform several natural language tasks — Link
DAIR.ai release new post which offers a light introduction to transfer learning for natural language processing and discusses key questions necessary to better understand this important research area — Link
MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has launched a research-industry collaboration focused on building financial technologies to create new business models and improve security, among other goals — Link
As part of the ML Education project, Google releases ML Learning Guides which are a set of tutorials to teach step-by-step machine learning essentials and best practices — Link
Matt Gardner releases new episode for the NLP Highlight podcast where he discusses with others about entity-based language models and how to use them to generate coherent stories — Link
On Code and Data…
The NLP-Progress repository helps to keep track of the recent progress in NLP research — Link
The Tensorflow team releases an intuitive guide to perform text classification using tf.keras — Link
Distill releases new paper entitled “Differentiable Image Parameterizations” which describes a powerful, under-explored tool for neural network visualizations and art (highly recommended) — Link
GluonNLP is a new tool for reproducing state-of-the-art research in NLP. It provides several implementations and features that makes it easier to build NLP-based prototypes and products — Link
Pythia is a modular framework that won the visual question answering challenge presented by FAIR — Link
On Industry…
Google, Microsoft, and Startups battle it out to become the leader of conversational bot technology — Link
Salesforce purchases Dataroma, an AI provider that benefits marketers, for $800 million — Link
alt Inc, a personalized artificial intelligence platform based in Tokyo, launches new distributed personal storage platform, al+ stack, to securely store personal data — Link
OpenAI announces their first batch of AI scholars — Link
Google AI announces BigQuery ML, a feature that allows data scientists to deploy machine learning models on large-scale structured and semi-structured datasets — Link
Quotes of the day…
An example of why autocomplete is still not perfect | read report here
Ian Goodfellow discusses troubling trends in machine learning scholarship
Worthy Mentions…
Here is a mini-course on Deep Learning with PyTorch (lecture slides and code included) — Link
Mu is a code editor for beginners to learn Python — Link
Nathan.ai state of AI report 2018 — Link
If you spot any errors or inaccuracies in this newsletter please comment below.
