Forward this message to
friends and colleagues
who want to stay up to
date with research and
news from the Allen
Institute for AI.
Subscribe here
AI2 NEWSLETTER | May
2020
In this
edition:
Updates on the
CORD-19 effort
and the new
tools being
built to explore
this important
resource, a demo
of the reasoning
capabilities of
language
transformers,
exciting new
state-of-the-art
QA work from the
Aristo team, and
more.
New Developments in Semantic
Scholar's CORD-19
resource
On March 16, our Semantic
Scholar team and collaborators
released the
COVID-19 Open Research
Dataset (CORD-19), along with a new challenge
for the research community in
partnership with Kaggle (learn
more in
this special edition of the
AI2 Newsletter). In the weeks since, hundreds
of organizations and thousands
of people have contributed to
the project, and the associated
Kaggle competition has become
the most popular ever on the
platform.
CORD-19 is a model for the power
of open science combined with
cutting-edge AI in the face of a
global crisis. Learn more in
this article from AI2 CEO
Oren Etzioni:
An NLP power
tool for
extractive
search over
CORD-19 using a
variety of query
modes. It allows
a level of query
expressivity and
control that is
substantially
more powerful
than existing
search
solutions.
Read the
blog post
from Yoav
Goldberg,
Research
Director at AI2
Israel, to learn
more.
SciSight is a
new framework of
tools for
visualizing the
emerging
literature
network around
COVID-19. Find
out what groups
are working on
what directions,
see how
biomedical
concepts
interact and
evolve over
time, and
discover new
connections.
Learn more in a
new blog
post from
the AI2
researchers
behind
SciSight.
The Aristo team
at AI2 released
a demo of their
new interactive
tool RuleTaker
that determines
whether
statements are
True or False
based on rules
provided in
natural
language. Try the
demo, and learn
more about the
research behind
RuleTaker:
UnifiedQA: state-of-the-art
results on six QA datasets
Researchers on the Aristo team
used the latest advances in
language modeling to build a
single pre-trained question
answering model, UnifiedQA, that
performs surprisingly well
across 17 QA datasets spanning 4
diverse formats (and achieves
state-of-the-art results on 6 of
those datasets!) –
try the demo, and learn more:
We ask the
Predoctoral
Young
Investigators on
the PRIOR team
about what
brought them to
AI2, what
experiences
they’ve had as
part of our
institute, and
where they’re
headed next:
Featured
AI2ers:
PRIOR's
Predoctoral
Young
Investigators
AI in the
Age of
COVID-19
AI is playing an
increasingly
vital role in
everything from
detection to
treatment to
prevention
during the
current
pandemic. Learn
about recent
developments,
future
potential, and
new concerns
in AI in the
Age of
COVID-19 by Oren
Etzioni and
Nicole DeCario.
