Update spaCy pipeline

40c4bf0 about 4 years ago

2.68 kB

tags:
  - spacy
language:
  - en
model-index:
  - name: en_stonk_pipeline
    results:
      - task:
          name: NER
          type: token-classification
        metrics:
          - name: NER Precision
            type: precision
            value: 0.8565043157
          - name: NER Recall
            type: recall
            value: 0.8348858173
          - name: NER F Score
            type: f_score
            value: 0.8455569081
      - task:
          name: TAG
          type: token-classification
        metrics:
          - name: TAG (XPOS) Accuracy
            type: accuracy
            value: 0.9726250474
      - task:
          name: UNLABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Unlabeled Attachment Score (UAS)
            type: f_score
            value: 0.9165718428
      - task:
          name: LABELED_DEPENDENCIES
          type: token-classification
        metrics:
          - name: Labeled Attachment Score (LAS)
            type: f_score
            value: 0.8978441095
      - task:
          name: SENTS
          type: token-classification
        metrics:
          - name: Sentences F-Score
            type: f_score
            value: 0.9038596962

pipeline to extract stonk names, need to adjust for general use as some stonk names are very short. Based on the standard spacy pipeline, but added a pipe and wanted to distribute it easily

Feature	Description
Name	`en_stonk_pipeline`
Version	`0.0.1`
spaCy	`>=3.4.1,<3.5.0`
Default Pipeline	`entity_ruler`
Components	`entity_ruler`
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) ClearNLP Constituent-to-Dependency Conversion (Emory University) WordNet 3.0 (Princeton University)
License	n/a
Author	FriendlyUser

Label Scheme

View label scheme (8 labels for 1 components)

Component	Labels
`entity_ruler`	`COMPANY`, `COUNTRY`, `DIVIDENDS`, `INDEX`, `MAYBE`, `STOCK`, `STOCK_EXCHANGE`, `THINGS`

Accuracy

Type	Score
`TOKEN_ACC`	99.93
`TOKEN_P`	99.57
`TOKEN_R`	99.58
`TOKEN_F`	99.57
`TAG_ACC`	97.26
`SENTS_P`	91.92
`SENTS_R`	88.90
`SENTS_F`	90.39
`DEP_UAS`	91.66
`DEP_LAS`	89.78
`ENTS_P`	85.65
`ENTS_R`	83.49
`ENTS_F`	84.56