Arabic Bert Embeddings (Large, Arabert Model, v2)

Description

Pretrained Bert Embeddings model, uploaded to Hugging Face, adapted and imported into Spark NLP. bert-large-arabertv2 is a Arabic model orginally trained by aubmindlab.

Download Copy S3 URI

How to use

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

tokenizer = Tokenizer() \
.setInputCols("document") \
.setOutputCol("token")

embeddings = BertEmbeddings.pretrained("bert_embeddings_bert_large_arabertv2","ar") \
.setInputCols(["document", "token"]) \
.setOutputCol("embeddings")

pipeline = Pipeline(stages=[documentAssembler, tokenizer, embeddings])

data = spark.createDataFrame([["أنا أحب شرارة 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 = BertEmbeddings.pretrained("bert_embeddings_bert_large_arabertv2","ar") 
.setInputCols(Array("document", "token")) 
.setOutputCol("embeddings")

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

val data = Seq("أنا أحب شرارة NLP").toDF("text")

val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("ar.embed.bert_large_arabertv2").predict("""أنا أحب شرارة NLP""")

Model Information

Model Name: bert_embeddings_bert_large_arabertv2
Compatibility: Spark NLP 3.4.2+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [bert]
Language: ar
Size: 1.4 GB
Case sensitive: true

References

  • https://huggingface.co/aubmindlab/bert-large-arabertv2
  • https://github.com/google-research/bert
  • https://arxiv.org/abs/2003.00104
  • https://github.com/WissamAntoun/pydata_khobar_meetup
  • http://alt.qcri.org/farasa/segmenter.html
  • /aubmindlab/bert-large-arabertv2/resolve/main/(https://github.com/google-research/bert/blob/master/multilingual.md)
  • https://github.com/elnagara/HARD-Arabic-Dataset
  • https://www.aclweb.org/anthology/D15-1299
  • https://staff.aub.edu.lb/~we07/Publications/ArSentD-LEV_Sentiment_Corpus.pdf
  • https://github.com/mohamedadaly/LABR
  • http://curtis.ml.cmu.edu/w/courses/index.php/ANERcorp
  • https://github.com/husseinmozannar/SOQAL
  • https://github.com/aub-mind/arabert/blob/master/AraBERT/README.md
  • https://arxiv.org/abs/2003.00104v2
  • https://archive.org/details/arwiki-20190201
  • https://www.semanticscholar.org/paper/1.5-billion-words-Arabic-Corpus-El-Khair/f3eeef4afb81223df96575adadf808fe7fe440b4
  • https://www.aclweb.org/anthology/W19-4619
  • https://sites.aub.edu.lb/mindlab/
  • https://www.yakshof.com/#/
  • https://www.behance.net/rahalhabib
  • https://www.linkedin.com/in/wissam-antoun-622142b4/
  • https://twitter.com/wissam_antoun
  • https://github.com/WissamAntoun
  • https://www.linkedin.com/in/fadybaly/
  • https://twitter.com/fadybaly
  • https://github.com/fadybaly