wietsedv's picture
update README
0b1c313
metadata
language:
  - he
license: apache-2.0
library_name: transformers
tags:
  - part-of-speech
  - token-classification
datasets:
  - universal_dependencies
metrics:
  - accuracy
model-index:
  - name: xlm-roberta-base-ft-udpos28-he
    results:
      - task:
          type: token-classification
          name: Part-of-Speech Tagging
        dataset:
          type: universal_dependencies
          name: Universal Dependencies v2.8
        metrics:
          - type: accuracy
            name: English Test accuracy
            value: 76.6
          - type: accuracy
            name: Dutch Test accuracy
            value: 73.7
          - type: accuracy
            name: German Test accuracy
            value: 70.5
          - type: accuracy
            name: Italian Test accuracy
            value: 75.1
          - type: accuracy
            name: French Test accuracy
            value: 71.3
          - type: accuracy
            name: Spanish Test accuracy
            value: 74.5
          - type: accuracy
            name: Russian Test accuracy
            value: 80.3
          - type: accuracy
            name: Swedish Test accuracy
            value: 79.3
          - type: accuracy
            name: Norwegian Test accuracy
            value: 75.7
          - type: accuracy
            name: Danish Test accuracy
            value: 80.4
          - type: accuracy
            name: Low Saxon Test accuracy
            value: 42.6
          - type: accuracy
            name: Akkadian Test accuracy
            value: 24.1
          - type: accuracy
            name: Armenian Test accuracy
            value: 77
          - type: accuracy
            name: Welsh Test accuracy
            value: 62.3
          - type: accuracy
            name: Old East Slavic Test accuracy
            value: 66.2
          - type: accuracy
            name: Albanian Test accuracy
            value: 73.9
          - type: accuracy
            name: Slovenian Test accuracy
            value: 72.5
          - type: accuracy
            name: Guajajara Test accuracy
            value: 21.4
          - type: accuracy
            name: Kurmanji Test accuracy
            value: 74.2
          - type: accuracy
            name: Turkish Test accuracy
            value: 71.8
          - type: accuracy
            name: Finnish Test accuracy
            value: 80.5
          - type: accuracy
            name: Indonesian Test accuracy
            value: 80
          - type: accuracy
            name: Ukrainian Test accuracy
            value: 78.8
          - type: accuracy
            name: Polish Test accuracy
            value: 78.9
          - type: accuracy
            name: Portuguese Test accuracy
            value: 78.6
          - type: accuracy
            name: Kazakh Test accuracy
            value: 77.2
          - type: accuracy
            name: Latin Test accuracy
            value: 73.5
          - type: accuracy
            name: Old French Test accuracy
            value: 50.6
          - type: accuracy
            name: Buryat Test accuracy
            value: 45
          - type: accuracy
            name: Kaapor Test accuracy
            value: 11.2
          - type: accuracy
            name: Korean Test accuracy
            value: 60.2
          - type: accuracy
            name: Estonian Test accuracy
            value: 81.4
          - type: accuracy
            name: Croatian Test accuracy
            value: 77.9
          - type: accuracy
            name: Gothic Test accuracy
            value: 13.7
          - type: accuracy
            name: Swiss German Test accuracy
            value: 44.8
          - type: accuracy
            name: Assyrian Test accuracy
            value: 17
          - type: accuracy
            name: North Sami Test accuracy
            value: 24.8
          - type: accuracy
            name: Naija Test accuracy
            value: 41.6
          - type: accuracy
            name: Latvian Test accuracy
            value: 80.1
          - type: accuracy
            name: Chinese Test accuracy
            value: 60.5
          - type: accuracy
            name: Tagalog Test accuracy
            value: 79.2
          - type: accuracy
            name: Bambara Test accuracy
            value: 21.1
          - type: accuracy
            name: Lithuanian Test accuracy
            value: 81
          - type: accuracy
            name: Galician Test accuracy
            value: 76.1
          - type: accuracy
            name: Vietnamese Test accuracy
            value: 64.4
          - type: accuracy
            name: Greek Test accuracy
            value: 67.4
          - type: accuracy
            name: Catalan Test accuracy
            value: 71.5
          - type: accuracy
            name: Czech Test accuracy
            value: 77.7
          - type: accuracy
            name: Erzya Test accuracy
            value: 32
          - type: accuracy
            name: Bhojpuri Test accuracy
            value: 50.7
          - type: accuracy
            name: Thai Test accuracy
            value: 69.2
          - type: accuracy
            name: Marathi Test accuracy
            value: 81.6
          - type: accuracy
            name: Basque Test accuracy
            value: 76.2
          - type: accuracy
            name: Slovak Test accuracy
            value: 78
          - type: accuracy
            name: Kiche Test accuracy
            value: 23.6
          - type: accuracy
            name: Yoruba Test accuracy
            value: 17.5
          - type: accuracy
            name: Warlpiri Test accuracy
            value: 22.3
          - type: accuracy
            name: Tamil Test accuracy
            value: 82.1
          - type: accuracy
            name: Maltese Test accuracy
            value: 18
          - type: accuracy
            name: Ancient Greek Test accuracy
            value: 45.4
          - type: accuracy
            name: Icelandic Test accuracy
            value: 81
          - type: accuracy
            name: Mbya Guarani Test accuracy
            value: 22
          - type: accuracy
            name: Urdu Test accuracy
            value: 70.9
          - type: accuracy
            name: Romanian Test accuracy
            value: 76.5
          - type: accuracy
            name: Persian Test accuracy
            value: 75.4
          - type: accuracy
            name: Apurina Test accuracy
            value: 22.2
          - type: accuracy
            name: Japanese Test accuracy
            value: 39.4
          - type: accuracy
            name: Hungarian Test accuracy
            value: 65.8
          - type: accuracy
            name: Hindi Test accuracy
            value: 75.2
          - type: accuracy
            name: Classical Chinese Test accuracy
            value: 44.3
          - type: accuracy
            name: Komi Permyak Test accuracy
            value: 35
          - type: accuracy
            name: Faroese Test accuracy
            value: 70.8
          - type: accuracy
            name: Sanskrit Test accuracy
            value: 12.1
          - type: accuracy
            name: Livvi Test accuracy
            value: 52.2
          - type: accuracy
            name: Arabic Test accuracy
            value: 77.4
          - type: accuracy
            name: Wolof Test accuracy
            value: 24.4
          - type: accuracy
            name: Bulgarian Test accuracy
            value: 82.1
          - type: accuracy
            name: Akuntsu Test accuracy
            value: 17
          - type: accuracy
            name: Makurap Test accuracy
            value: 8.2
          - type: accuracy
            name: Kangri Test accuracy
            value: 39.9
          - type: accuracy
            name: Breton Test accuracy
            value: 56.7
          - type: accuracy
            name: Telugu Test accuracy
            value: 81.4
          - type: accuracy
            name: Cantonese Test accuracy
            value: 57.7
          - type: accuracy
            name: Old Church Slavonic Test accuracy
            value: 40.3
          - type: accuracy
            name: Karelian Test accuracy
            value: 60
          - type: accuracy
            name: Upper Sorbian Test accuracy
            value: 61.2
          - type: accuracy
            name: South Levantine Arabic Test accuracy
            value: 64.5
          - type: accuracy
            name: Komi Zyrian Test accuracy
            value: 29
          - type: accuracy
            name: Irish Test accuracy
            value: 58.7
          - type: accuracy
            name: Nayini Test accuracy
            value: 41
          - type: accuracy
            name: Munduruku Test accuracy
            value: 9.5
          - type: accuracy
            name: Manx Test accuracy
            value: 21.8
          - type: accuracy
            name: Skolt Sami Test accuracy
            value: 27.2
          - type: accuracy
            name: Afrikaans Test accuracy
            value: 73.3
          - type: accuracy
            name: Old Turkish Test accuracy
            value: 43.4
          - type: accuracy
            name: Tupinamba Test accuracy
            value: 21.9
          - type: accuracy
            name: Belarusian Test accuracy
            value: 78.5
          - type: accuracy
            name: Serbian Test accuracy
            value: 78.9
          - type: accuracy
            name: Moksha Test accuracy
            value: 29.7
          - type: accuracy
            name: Western Armenian Test accuracy
            value: 69.6
          - type: accuracy
            name: Scottish Gaelic Test accuracy
            value: 51.3
          - type: accuracy
            name: Khunsari Test accuracy
            value: 36.5
          - type: accuracy
            name: Hebrew Test accuracy
            value: 93.8
          - type: accuracy
            name: Uyghur Test accuracy
            value: 70.2
          - type: accuracy
            name: Chukchi Test accuracy
            value: 27.1

XLM-RoBERTa base Universal Dependencies v2.8 POS tagging: Hebrew

This model is part of our paper called:

  • Make the Best of Cross-lingual Transfer: Evidence from POS Tagging with over 100 Languages

Check the Space for more details.

Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification

tokenizer = AutoTokenizer.from_pretrained("wietsedv/xlm-roberta-base-ft-udpos28-he")
model = AutoModelForTokenClassification.from_pretrained("wietsedv/xlm-roberta-base-ft-udpos28-he")