zh_core_web_sm / README.md
adrianeboyd's picture
Update spaCy pipeline
e24daa3
metadata
tags:
  - spacy
  - token-classification
language:
  - zh
license: mit
model-index:
  - name: zh_core_web_sm
    results:
      - task:
          name: NER
          type: token-classification
        metrics:
          - name: NER Precision
            type: precision
            value: 0.7203462148
          - name: NER Recall
            type: recall
            value: 0.6493406593
          - name: NER F Score
            type: f_score
            value: 0.6830029475
      - task:
          name: TAG
          type: token-classification
        metrics:
          - name: TAG (XPOS) Accuracy
            type: accuracy
            value: 0.8933253054
      - task:
          name: UNLABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Unlabeled Attachment Score (UAS)
            type: f_score
            value: 0.6960047338
      - task:
          name: LABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Labeled Attachment Score (LAS)
            type: f_score
            value: 0.640776699
      - task:
          name: SENTS
          type: token-classification
        metrics:
          - name: Sentences F-Score
            type: f_score
            value: 0.7514211886

Details: https://spacy.io/models/zh#zh_core_web_sm

Chinese pipeline optimized for CPU. Components: tok2vec, tagger, parser, senter, ner, attribute_ruler.

Feature Description
Name zh_core_web_sm
Version 3.7.0
spaCy >=3.7.0,<3.8.0
Default Pipeline tok2vec, tagger, parser, attribute_ruler, ner
Components tok2vec, tagger, parser, senter, attribute_ruler, ner
Vectors 0 keys, 0 unique vectors (0 dimensions)
Sources OntoNotes 5 (Ralph Weischedel, Martha Palmer, Mitchell Marcus, Eduard Hovy, Sameer Pradhan, Lance Ramshaw, Nianwen Xue, Ann Taylor, Jeff Kaufman, Michelle Franchini, Mohammed El-Bachouti, Robert Belvin, Ann Houston)
CoreNLP Universal Dependencies Converter (Stanford NLP Group)
License MIT
Author Explosion

Label Scheme

View label scheme (100 labels for 3 components)
Component Labels
tagger AD, AS, BA, CC, CD, CS, DEC, DEG, DER, DEV, DT, ETC, FW, IJ, INF, JJ, LB, LC, M, MSP, NN, NR, NT, OD, ON, P, PN, PU, SB, SP, URL, VA, VC, VE, VV, X, _SP
parser ROOT, acl, advcl:loc, advmod, advmod:dvp, advmod:loc, advmod:rcomp, amod, amod:ordmod, appos, aux:asp, aux:ba, aux:modal, aux:prtmod, auxpass, case, cc, ccomp, compound:nn, compound:vc, conj, cop, dep, det, discourse, dobj, etc, mark, mark:clf, name, neg, nmod, nmod:assmod, nmod:poss, nmod:prep, nmod:range, nmod:tmod, nmod:topic, nsubj, nsubj:xsubj, nsubjpass, nummod, parataxis:prnmod, punct, xcomp
ner CARDINAL, DATE, EVENT, FAC, GPE, LANGUAGE, LAW, LOC, MONEY, NORP, ORDINAL, ORG, PERCENT, PERSON, PRODUCT, QUANTITY, TIME, WORK_OF_ART

Accuracy

Type Score
TOKEN_ACC 95.85
TOKEN_P 94.58
TOKEN_R 91.36
TOKEN_F 92.94
TAG_ACC 89.33
SENTS_P 77.85
SENTS_R 72.62
SENTS_F 75.14
DEP_UAS 69.60
DEP_LAS 64.08
ENTS_P 72.03
ENTS_R 64.93
ENTS_F 68.30