Downloads · 30 days
8
1% of all-time downloads
hiroshi-matsuda-rit/ja_gsd_bert_wwm_unidic_lite
ja_gsd_bert_wwm_unidic_lite is a token classification model from hiroshi-matsuda-rit. Use it when you need labels on individual words, such as names. It is set up for spacy.
Japanese transformer pipeline (bert-base). Components: transformer, parser, ner.
Downloads · 30 days
8
1% of all-time downloads
All-time downloads
582
Public
Repo size
1.3 GB
Likes
0
Public
Click a slice to open those files.
.bin443 MB · 52%
From the Hugging Face model README
Japanese transformer pipeline (bert-base). Components: transformer, parser, ner.
| Feature | Description |
|---|---|
| Name | ja_gsd_bert_wwm_unidic_lite |
| Version | 3.1.1 |
| spaCy | >=3.1.0,<3.2.0 |
| Default Pipeline | transformer, parser, ner |
| Components | transformer, parser, ner |
| Vectors | 0 keys, 0 unique vectors (0 dimensions) |
| Sources | UD_Japanese-GSD<br />UD_Japanese-GSD r2.8+NE<br />SudachiDict_core<br />cl-tohoku/bert-base-japanese-whole-word-masking<br />unidic_lite |
| License | CC BY-SA 4.0 |
| Author | Megagon Labs Tokyo. |
| Component | Labels |
|---|---|
parser | ROOT, acl, advcl, advmod, amod, aux, case, cc, ccomp, compound, cop, csubj, dep, det, dislocated, fixed, mark, nmod, nsubj, nummod, obj, obl, punct |
ner | CARDINAL, DATE, EVENT, FAC, GPE, LANGUAGE, LAW, LOC, MONEY, MOVEMENT, NORP, ORDINAL, ORG, PERCENT, PERSON, PET_NAME, PHONE, PRODUCT, QUANTITY, TIME, TITLE_AFFIX, WORK_OF_ART |
| Type | Score |
|---|---|
DEP_UAS | 93.68 |
DEP_LAS | 92.61 |
SENTS_P | 92.02 |
SENTS_R | 95.46 |
SENTS_F | 93.71 |
ENTS_F | 84.04 |
ENTS_P | 84.96 |
ENTS_R | 83.14 |
TAG_ACC | 0.00 |
TRANSFORMER_LOSS | 28861.67 |
PARSER_LOSS | 1306248.63 |
NER_LOSS | 13993.36 |