Bangla Bert Embeddings (from Kowsher)

Description

Pretrained Bert Embeddings model, uploaded to Hugging Face, adapted and imported into Spark NLP. bangla-bert is a Bangla model orginally trained by Kowsher.

Download

How to use

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

tokenizer = Tokenizer() \
    .setInputCols("document") \
    .setOutputCol("token")
  
embeddings = BertEmbeddings.pretrained("bert_embeddings_bangla_bert","bn") \
    .setInputCols(["document", "token"]) \
    .setOutputCol("embeddings")
    
pipeline = Pipeline(stages=[documentAssembler, tokenizer, embeddings])

data = spark.createDataFrame([["আমি স্পার্ক এনএলপি ভালোবাসি"]]).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 = BertEmbeddings.pretrained("bert_embeddings_bangla_bert","bn") 
    .setInputCols(Array("document", "token")) 
    .setOutputCol("embeddings")

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

val data = Seq("আমি স্পার্ক এনএলপি ভালোবাসি").toDF("text")

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

Model Information

Model Name: bert_embeddings_bangla_bert
Compatibility: Spark NLP 3.4.2+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [bert]
Language: bn
Size: 615.0 MB
Case sensitive: true

References

  • https://huggingface.co/Kowsher/bangla-bert
  • https://github.com/Kowsher/bert-base-bangla
  • https://arxiv.org/abs/1810.04805
  • https://github.com/google-research/bert
  • https://www.kaggle.com/gakowsher/bangla-language-model-dataset
  • https://ssrn.com/abstract=
  • http://kowsher.org/