osanseviero's picture
osanseviero HF staff
Update spaCy pipeline
3d76675
metadata
tags:
  - spacy
  - token-classification
language:
  - ja
license: CC-BY-SA-4.0
model-index:
  - name: ja_gsd_bert_wwm_unidic_lite
    results:
      - task:
          name: NER
          type: token-classification
        metrics:
          - name: NER Precision
            type: precision
            value: 0.8496143959
          - name: NER Recall
            type: recall
            value: 0.8314465409
          - name: NER F Score
            type: f_score
            value: 0.840432295
      - task:
          name: POS
          type: token-classification
        metrics:
          - name: POS Accuracy
            type: accuracy
            value: 0
      - task:
          name: SENTER
          type: token-classification
        metrics:
          - name: SENTER Precision
            type: precision
            value: 0.9201520913
          - name: SENTER Recall
            type: recall
            value: 0.9546351085
          - name: SENTER F Score
            type: f_score
            value: 0.9370764763
      - task:
          name: UNLABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Unlabeled Dependencies Accuracy
            type: accuracy
            value: 0.9367795389
      - task:
          name: LABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Labeled Dependencies Accuracy
            type: accuracy
            value: 0.9367795389

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
UD_Japanese-GSD r2.8+NE
SudachiDict_core
cl-tohoku/bert-base-japanese-whole-word-masking
unidic_lite
License CC BY-SA 4.0
Author Megagon Labs Tokyo.

Label Scheme

View label scheme (45 labels for 2 components)
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

Accuracy

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