README.md

metadata

language:
  - mt
datasets:
  - MLRS/korpus_malti
model-index:
  - name: mBERTu
    results:
      - task:
          type: dependency-parsing
          name: Dependency Parsing
        dataset:
          type: universal_dependencies
          args: mt_mudt
          name: Maltese Universal Dependencies Treebank (MUDT)
        metrics:
          - type: uas
            value: 92.1
            name: Unlabelled Attachment Score
          - type: las
            value: 87.87
            name: Labelled Attachment Score
      - task:
          type: part-of-speech-tagging
          name: Part-of-Speech Tagging
        dataset:
          type: mlrs_pos
          name: MLRS POS dataset
        metrics:
          - type: accuracy
            value: 98.66
            name: UPOS Accuracy
            args: upos
          - type: accuracy
            value: 98.58
            name: XPOS Accuracy
            args: xpos
      - task:
          type: named-entity-recognition
          name: Named Entity Recognition
        dataset:
          type: wikiann
          name: WikiAnn (Maltese)
          args: mt
        metrics:
          - type: f1
            args: span
            value: 86.6
            name: Span-based F1
      - task:
          type: sentiment-analysis
          name: Sentiment Analysis
        dataset:
          type: mt-sentiment-analysis
          name: Maltese Sentiment Analysis Dataset
        metrics:
          - type: f1
            args: macro
            value: 76.79
            name: Macro-averaged F1
widget:
  - text: Malta huwa pajjiż fl-[MASK].

mBERTu

A Maltese multilingual model pre-trained on the Korpus Malti v4.0 using multilingual BERT as the initial checkpoint.