hiroshi-matsuda-rit commited on
Commit
1b72ad3
1 Parent(s): ebeb9d2

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +70 -1
README.md CHANGED
@@ -1,3 +1,72 @@
1
  ---
2
- license: mit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ language:
3
+ - ja
4
+ thumbnail: "https://raw.githubusercontent.com/megagonlabs/ginza/static/docs/images/GiNZA_logo_4c_s.png"
5
+ tags:
6
+ - PyTorch
7
+ - Transformers
8
+ - spaCy
9
+ - ELECTRA
10
+ - GiNZA
11
+ - mC4
12
+ - UD_Japanese-BCCWJ
13
+ - GSK2014-A
14
+ - ja
15
+ - MIT
16
+ license: "mit"
17
+ datasets:
18
+ - mC4
19
+ - UD_Japanese_BCCWJ-r2.8
20
+ - GSK2014-A(2019)
21
+ metrics:
22
+ - UAS
23
+ - LAS
24
+ - UPOS
25
  ---
26
+
27
+ # transformers-ud-japanese-electra-ginza-520 (sudachitra-wordpiece, mC4 Japanese)
28
+
29
+ This is an [ELECTRA](https://github.com/google-research/electra) model pretrained on approximately 200M Japanese sentences extracted from the [mC4](https://huggingface.co/datasets/mc4) and finetuned by [spaCy v3](https://spacy.io/usage/v3) on [UD\_Japanese\_BCCWJ r2.8](https://universaldependencies.org/treebanks/ja_bccwj/index.html).
30
+
31
+ The base pretrain model is [megagonlabs/transformers-ud-japanese-electra-base-discrimininator](https://huggingface.co/megagonlabs/transformers-ud-japanese-electra-base-discriminator).
32
+
33
+ The entire spaCy v3 model is distributed as a python package named [`ja_ginza_electra`](https://pypi.org/project/ja-ginza-electra/) from PyPI along with [`GiNZA v5`](https://github.com/megagonlabs/ginza) which provides some custom pipeline components to recognize the Japanese bunsetu-phrase structures.
34
+ Try running it as below:
35
+ ```console
36
+ $ pip install ginza ja_ginza_electra
37
+ $ ginza
38
+ ```
39
+
40
+ ## Licenses
41
+
42
+ The models are distributed under the terms of the [MIT License](https://opensource.org/licenses/mit-license.php).
43
+
44
+ ## Acknowledgments
45
+
46
+ This model is permitted to be published under the `MIT License` under a joint research agreement between NINJAL (National Institute for Japanese Language and Linguistics) and Megagon Labs Tokyo.
47
+
48
+ ## Citations
49
+ - [mC4](https://huggingface.co/datasets/mc4)
50
+
51
+ Contains information from `mC4` which is made available under the [ODC Attribution License](https://opendatacommons.org/licenses/by/1-0/).
52
+ ```
53
+ @article{2019t5,
54
+ author = {Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu},
55
+ title = {Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer},
56
+ journal = {arXiv e-prints},
57
+ year = {2019},
58
+ archivePrefix = {arXiv},
59
+ eprint = {1910.10683},
60
+ }
61
+ ```
62
+
63
+ - [UD\_Japanese\_BCCWJ r2.8](https://universaldependencies.org/treebanks/ja_bccwj/index.html)
64
+
65
+ ```
66
+ Asahara, M., Kanayama, H., Tanaka, T., Miyao, Y., Uematsu, S., Mori, S.,
67
+ Matsumoto, Y., Omura, M., & Murawaki, Y. (2018).
68
+ Universal Dependencies Version 2 for Japanese.
69
+ In LREC-2018.
70
+ ```
71
+
72
+ - [GSK2014-A(2019)](https://www.gsk.or.jp/catalog/gsk2014-a/)