News

News 2021-04-16

Invited as keynote speaker at AWS Summit Online Japan

Yamada will talk about developing AI that understands documents


News 2021-03-15

Studio Ousia AI “Sōseki” enters open beta

Live demos showcasing at SXSW 2021


News 2021-03-04

New AI model Sōseki will be showcased at SXSW

We will be exhibiting as a Todai-to-Texas program team


News 2020-12-16

Our new model Sōseki ranked 2nd in a track of NeurIPS Efficient QA competition

Sōseki had 62% accuracy according to human evaluation


News 2020-10-17

Two papers were accepted at EMNLP

Our two papers were accepted at the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)


News 2020-04-30

New natural language AI model (LUKE) tops SQuAD v1.1 dataset leaderboard

Becoming the best model in one of the most renown question answering dataset


News 2020-04-28

Tool to semi-automate training data creation for AI “SMART ANNOTATOR” launched

Text-clustering tasks and training data creation for AIs made easy


News 2019-09-02

Paper co-authored with the University of Maryland team accepted at TACL

A paper on the development of an adversarial question answering dataset was accepted by TACL.


News 2019-09-02

Paper accepted at The SIGNLL Conference on Computational Natural Language Learning

Paper proposing a novel text classification method using neural network with Wikipedia entities as inputs was accepted at The CoNLL.


News 2017-12-14

Studio Ousia’s AI Defeats a Team of Human Quiz Champions

Studio Ousia, the AI quiz champion at NIPS 2017 (the top machine learning conference), also defeats humans.