LINGUIST List 32.1030

Sun Mar 21 2021

Software: Releasing Python-based Dependency Parser and Models for UD

Editor for this issue: Everett Green <everettlinguistlist.org>



Date: 18-Mar-2021
From: Annemarie Friedrich <annemarie.friedrichgmail.com>
Subject: Releasing Python-based Dependency Parser and Models for UD
E-mail this message to a friend

Dear all,

we are releasing STEPS, a graph-based parser implemented in Python for basic and enhanced Universal Dependencies. STEPS currently obtains state-of-the-art results on UD treebanks of Arabic, Chinese, Czech, English, Finnish, German, Hindi, Italian, Latvian, and Russian. (More details in our Arxiv pre-print here: https://arxiv.org/pdf/2010.12699.pdf)
STEPS is free software released under the AGPLv3 license. The source code, as well as download instructions for pre-trained models for 12 languages, is available at https://github.com/boschresearch/steps-parser.

We welcome feedback regarding the usability and performance of the system.

Best Regards,
Stefan Grünewald
(University of Stuttgart & Bosch Center for Artificial Intelligence)

Linguistic Field(s): Computational Linguistics

Subject Language(s): Arabana (ard)


Page Updated: 21-Mar-2021