LINGUIST List 32.2724

Fri Aug 20 2021

Confs: Computational Linguistics/USA

Editor for this issue: Everett Green <>

Date: 18-Aug-2021
From: Carl Rubino <>
Subject: Open Automatic Speech Recognition 2021 (OpenASR21) Challenge
E-mail this message to a friend

Open Automatic Speech Recognition 2021 (OpenASR21) Challenge
Short Title: OpenASR21

Date: 30-Aug-2021 - 02-Nov-2021
Location: Virtual, USA
Contact: Carl Rubino
Contact Email: < click here to access email >
Meeting URL:

Linguistic Field(s): Computational Linguistics

Subject Language(s): Amharic; Chinese, Yue; English; GuaranĂ­, Paraguayan; Kurdish, Central; Mongolian, Middle; Pashto; Persian, Iranian; Swahili; Tagalog; Tamil; Vietnamese

Meeting Description:

NIST and IARPA are pleased to announce our latest Automatic Speech Recognition Challenge, OpenASR21, open to all. This new challenge was created out of the IARPA MATERIAL Program, which encompasses more NLP tasks including CLIR (cross-language information retrieval), machine translation, domain classification, and cross-language summarization. To kickstart work in this new challenge, NIST will distribute ASR training sets to all registered participants for 15 lower resourced languages, to include all ten languages from last year's OpenASR20 challenge as well as five new languages (Amharic, Cantonese, Georgian, Farsi, Guarani, Javanese, Kazakh, Kurmanji Kurdish, Mongolian, Pashto, Somali, Swahili, Tagalog, Tamil, Vietnamese). New to OpenASR21 will be case sensitive scoring on three languages (Kazakh, Swahili and Tagalog) to help assess performance on proper nouns, and three different training conditions detailed in the Evaluation Plan.

Please find the current evaluation plan, toolkit and other resources, and link to registration here:

Please note that the conference dates provided here are the dates for which the NIST server will accept development submissions. The Evaluation Period will be Nov 3-10. Challenge winners will be announced on the NIST website.

Page Updated: 20-Aug-2021