Italian DistilBERT Embeddings (from indigo-ai)

Description

Pretrained DistilBERT Embeddings model, uploaded to Hugging Face, adapted and imported into Spark NLP. BERTino is a Italian model orginally trained by indigo-ai.

Download

How to use

documentAssembler = DocumentAssembler() \
    .setInputCol("text") \
    .setOutputCol("document")

tokenizer = Tokenizer() \
    .setInputCols("document") \
    .setOutputCol("token")
  
embeddings = DistilBertEmbeddings.pretrained("distilbert_embeddings_BERTino","it") \
    .setInputCols(["document", "token"]) \
    .setOutputCol("embeddings")
    
pipeline = Pipeline(stages=[documentAssembler, tokenizer, embeddings])

data = spark.createDataFrame([["Adoro Spark NLP"]]).toDF("text")

result = pipeline.fit(data).transform(data)
val documentAssembler = new DocumentAssembler() 
      .setInputCol("text") 
      .setOutputCol("document")
 
val tokenizer = new Tokenizer() 
    .setInputCols(Array("document"))
    .setOutputCol("token")

val embeddings = DistilBertEmbeddings.pretrained("distilbert_embeddings_BERTino","it") 
    .setInputCols(Array("document", "token")) 
    .setOutputCol("embeddings")

val pipeline = new Pipeline().setStages(Array(documentAssembler, tokenizer, embeddings))

val data = Seq("Adoro Spark NLP").toDF("text")

val result = pipeline.fit(data).transform(data)

Model Information

Model Name: distilbert_embeddings_BERTino
Compatibility: Spark NLP 3.4.2+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [bert]
Language: it
Size: 253.3 MB
Case sensitive: true

References

  • https://huggingface.co/indigo-ai/BERTino
  • https://indigo.ai/en/
  • https://www.corpusitaliano.it/
  • https://corpora.dipintra.it/public/run.cgi/corp_info?corpname=itwac_full
  • https://universaldependencies.org/treebanks/it_partut/index.html
  • https://universaldependencies.org/treebanks/it_isdt/index.html
  • https://figshare.com/articles/Learning_multilingual_named_entity_recognition_from_Wikipedia/5462500