Share this newsletter with friends and colleagues who want to stay up to date with research and events at the Allen Institute for Artificial Intelligence. Subscribe here .

AI2 NEWSLETTER | September 2019

Aristo Aces 8th-Grade Science

This month we unveiled the latest version of Aristo, an AI system capable of scoring over 90% on an 8th-grade multiple-choice science exam.

This achievement is the conclusion of Aristo's multi-year endeavor to successfully tackle grade school science exams, and the team now plans to shift focus to exciting new challenges in the field of machine reasoning. Check out this interview with Aristo Research Lead Peter Clark, and learn more in this article:

A Breakthrough for A.I. Technology: Passing an 8th-Grade Science Test

New York Times [page B1, print edition]

The Aristo team at AI2

Standing, left to right: Dirk Groeneveld, Kyle Richardson, Ashish Sabharwal, Peter Clark, Oren Etzioni, Oyvind Tafjord, Carissa Schoenick, Bhavana Dalvi, Tushar Khot.
Seated: Michael Schmitz, Daniel Khashabi, Niket Tandon, Sumithra Bhakthavatsalam, Michal Guerquin

Discover Supplement-Drug Interactions

AI2 has released a new tool called Supp.AI that searches our AI-curated corpus of millions of biomedical papers and identifies potential supplement/drug interactions noted in the scientific literature.

Dietary and herbal supplements are popular but unregulated, and supplements can interact or interfere with the action of prescription or over-the-counter medications — Supp.AI is a free tool for users to find accurate and timely scientific evidence for these potentially dangerous interactions, and discuss them with their doctors.

Learn more in the related coverage:

SuppAI Uses Machine Learning to Identify Supplement Interactions —VentureBeat
How Do Drugs Interact with Supplements? Supp.AI Search Engine Tracks Down the Clues — GeekWire
Is that Supplement Safe to Take? This AI Tool Scours Research to Find Out — TechCrunch
Free online tool identifies dangerous drug/supplement combinations — KING5

AllenNLP Summit

In late August we held the first AllenNLP Summit, where users from across the country joined us for a full day working directly with our team and providing us valuable feedback on our existing ideas as well as new directions to consider. Learn more about the summit and what's next for AllenNLP in our blog post!

NEW DATASETS

Abductive Commonsense Reasoning

Chandra Bhagavatula, Ronan Le Bras, Chaitanya Malaviya, Keisuke Sakaguchi, Ari Holtzman, Hannah Rashkin, Doug Downey, Wen-tau Yih, Yejin Choi

We conceptualize a new task of Abductive NLI and introduce a challenge dataset, ART, that consists of over 20k commonsense narrative contexts and 200k explanations, formulated as multiple choice questions for easy automatic evaluation. Check out the new AI2 Leaderboard for this dataset.

Cosmos QA: Machine Reading Comprehension with Contextual Commonsense Reasoning

Lifu Huang, Ronan Le Bras, Chandra Bhagavatula, Yejin Choi

Cosmos QA is a large-scale dataset of 35.6K problems that require commonsense-based reading comprehension, formulated as multiple-choice questions. It focuses on reading between the lines over a diverse collection of people's everyday narratives, asking questions about the likely causes or effects of events that require reasoning beyond the exact text spans in the context. Visit the new Cosmos QA Leaderboard here .

Quoref: A Reading Comprehension Dataset with Questions Requiring Coreferential Reasoning

Pradeep Dasigi, Nelson F. Liu, Ana Marasovi'c, Noah A. Smith, Matt Gardner

Quoref is a QA dataset that tests the coreferential reasoning capability of reading comprehension systems. In this span-selection benchmark containing 24K questions over 4.7K paragraphs from Wikipedia, a system must resolve hard coreferences before selecting the appropriate span(s) in the paragraphs for answering questions. Check out the new AI2 Leaderboard for this dataset and try Quoref yourself!

Check out all of our leaderboards at leaderboard.allenai.org →