LINSPECTOR

Beta-version (Currently under test)

Language Inspector

LINSPECTOR is a multilingual inspector to analyze word representations of your pre-trained AllenNLP models, HuggingFace's Transformers models or static embeddings for 52 languages.

Get Started Demo Video Play
File Types

LINSPECTOR supports a number of AllenNLP Transformers models, HuggingFace's Transformers models and static embeddings files. Learn more.

  • AllenNLP Archive .tar.gz
    • BiaffineDependencyParser
    • CrfTagger
    • ESIM
    • SimpleTagger
  • HuggingFace's Transformers (PyTorch) .zip
    • ALBERT
    • BERT
    • CTRL
    • CamemBERT
    • DistilBERT
    • OpenAI GPT
    • OpenAI GPT2
    • RoBERTa
    • Transformer-XL
    • XLM
    • XLNet
  • Static Embeddings .vec
Languages 52*

LINSPECTOR provides probing tasks for a diverse set of languages. Every language has a different set of available tasks.

  • Afrikaans*
  • Ancient Greek*
  • Arabic
  • Armenian
  • Bambara*
  • Basque
  • Bulgarian
  • Catalan
  • Chinese*
  • Coptic*
  • Croatian*
  • Czech
  • Danish
  • Dutch
  • English
  • Estonian
  • Finnish
  • French
  • Galician*
  • German
  • Gothic*
  • Greek
  • Hebrew*
  • Hindi*
  • Hungarian
  • Indonesian*
  • Irish*
  • Italian
  • Japanese*
  • Korean*
  • Latin*
  • Latvian*
  • Macedonian
  • Northern Sami*
  • Norwegian*
  • Old Church Slavonic*
  • Persian*
  • Polish
  • Portuguese
  • Quechuan
  • Romanian
  • Russian
  • Serbian
  • Serbo-Croatian
  • Slovak*
  • Slovenian*
  • Spanish
  • Swedish
  • Turkish
  • Ukrainian*
  • Urdu*
  • Vietnamese
  • * The marked languages are only supported by the contextual probing tasks. 49 languages are supported.
  • ⁺ The marked languages are only supported by the static probing tasks. 28 languages are supported.
Probing Tasks 22*

Probing tasks analyze pre-trained embeddings for a single linguistic feature such as case marking, gender, person, or tense. Learn more.

  • Animacy*
  • Aspect*
  • Case Marking
  • Comparison
  • Definiteness*
  • Evidentiality*
  • Finiteness*
  • Gender
  • Mood
  • Morphological Feature Count
  • Number
  • Odd Morphological Feature
  • Person
  • Polarity
  • POS
  • Possession
  • Pseudoword
  • Shared Morphological Feature
  • Tense
  • Valency*
  • Voice
  • Word Length
  • * The marked probing tasks are only available in the contextual dataset. 19 probing tasks are contextual.
  • ⁺ The marked probing tasks are only available in the static dataset. 16 probing tasks are static.