Skip to content

Latest commit

 

History

History
82 lines (56 loc) · 3.09 KB

README.md

File metadata and controls

82 lines (56 loc) · 3.09 KB

MorphoRuEval

materials for MorphoRuEval-2017 track

http://www.dialog-21.ru/evaluation/2017/morphorueval/

alt text

Results of the tracks

Resulting team table:

https://docs.google.com/spreadsheets/d/1npLGIvfxtjRiLRuQjd1rkbnr-nlKBWC_yKd_0nRYTnE/edit

Resulting open source technologies and methods

OpenSource Tools

Papers

Presentations

Test set, scripts for its extraction and the best tries of all the teams (by litera):

https://drive.google.com/drive/folders/0B600DBw1ZmZASDFRVkJVd0pqNXM

Morphological standard and rules:

https://github.com/dialogue-evaluation/morphoRuEval-2017/blob/master/morphostandard

illustration.txt - file with examples of format and data tagging

DET.txt - a closed list of all determiners for Russian in Universal Dependencies format

PRON.txt - a closed list of all pronouns for Russian in Universal Dependencies format

https://github.com/kmike/dialog2017 scripts to unify the data format to json or conllu

Training data:

General Internet-Corpus of Russian UD

https://github.com/dialogue-evaluation/morphoRuEval-2017/blob/master/GIKRYA_texts.rar

Russian National Corpus UD

(please sign the license!) https://github.com/dialogue-evaluation/morphoRuEval-2017/blob/master/RNC_license_1mln-UD.pdf

https://github.com/dialogue-evaluation/morphoRuEval-2017/blob/master/RNC_texts.rar

OpenCorpora UD

https://github.com/dialogue-evaluation/morphoRuEval-2017/blob/master/OpenCorpora_Texts.rar

Plain text materials:

Live Journal from General Internet-Corpus of Russian, 30 million words:

https://github.com/dialogue-evaluation/morphoRuEval-2017/tree/LiveJournal

Archives have no password

Librusec, 300 million words:

https://github.com/dialogue-evaluation/morphoRuEval-2017/tree/librusec

Password - Morphorueval

Social networks, 50 million words:

https://github.com/dialogue-evaluation/morphoRuEval-2017/tree/social_media

(Twitter, VKontakte and Facebook) Archives have no password