mk_core_news_md / README.md
adrianeboyd's picture
Update spaCy pipeline
470f9da
metadata
tags:
  - spacy
  - token-classification
language:
  - mk
license: cc-by-sa-4.0
model-index:
  - name: mk_core_news_md
    results:
      - task:
          name: NER
          type: token-classification
        metrics:
          - name: NER Precision
            type: precision
            value: 0.7472245944
          - name: NER Recall
            type: recall
            value: 0.7446808511
          - name: NER F Score
            type: f_score
            value: 0.7459505541
      - task:
          name: POS
          type: token-classification
        metrics:
          - name: POS (UPOS) Accuracy
            type: accuracy
            value: 0.9260857837
      - task:
          name: UNLABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Unlabeled Attachment Score (UAS)
            type: f_score
            value: 0.6771344455
      - task:
          name: LABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Labeled Attachment Score (LAS)
            type: f_score
            value: 0.5201177625
      - task:
          name: SENTS
          type: token-classification
        metrics:
          - name: Sentences F-Score
            type: f_score
            value: 0.7323943662

Details: https://spacy.io/models/mk#mk_core_news_md

Macedonian pipeline optimized for CPU. Components: tok2vec, morphologizer, parser, senter, ner, attribute_ruler, lemmatizer.

Feature Description
Name mk_core_news_md
Version 3.7.0
spaCy >=3.7.0,<3.8.0
Default Pipeline morphologizer, parser, attribute_ruler, lemmatizer, ner
Components morphologizer, parser, senter, attribute_ruler, lemmatizer, ner
Vectors 274587 keys, 20000 unique vectors (300 dimensions)
Sources Macedonian Corpus (Damjan Zlatinov, Melanija Gerasimovska, Borijan Georgievski, Marija Todosovska)
spaCy lookups data (Explosion)
Explosion fastText Vectors (cbow, OSCAR Common Crawl + Wikipedia) (Explosion)
License CC BY-SA 4.0
Author Explosion

Label Scheme

View label scheme (54 labels for 3 components)
Component Labels
morphologizer POS=PROPN, POS=AUX, POS=ADJ, POS=NOUN, POS=ADP, POS=PUNCT, POS=CONJ, POS=NUM, POS=VERB, POS=PRON, POS=ADV, POS=SCONJ, POS=PART, POS=SYM, _, POS=SPACE, POS=X, POS=INTJ
parser ROOT, advmod, att, aux, cc, dep, det, dobj, iobj, neg, nsubj, pobj, poss, pozm, pozv, prep, punct, relcl
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 100.00
TOKEN_P 100.00
TOKEN_R 100.00
TOKEN_F 100.00
SENTS_P 80.00
SENTS_R 67.53
SENTS_F 73.24
DEP_UAS 67.71
DEP_LAS 52.01
ENTS_P 74.72
ENTS_R 74.47
ENTS_F 74.60
POS_ACC 92.61