Ukrainian RoBERTa Embeddings

Description

Pretrained RoBERTa Embeddings model, uploaded to Hugging Face, adapted and imported into Spark NLP. ukr-roberta-base is a Ukrainian model orginally trained by youscan.

Download

How to use

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

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

embeddings = RoBertaEmbeddings.pretrained("roberta_embeddings_ukr_roberta_base","uk") \
.setInputCols(["document", "token"]) \
.setOutputCol("embeddings")

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

data = spark.createDataFrame([["Я люблю 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 = RoBertaEmbeddings.pretrained("roberta_embeddings_ukr_roberta_base","uk") 
.setInputCols(Array("document", "token")) 
.setOutputCol("embeddings")

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

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

val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("uk.embed.ukr_roberta_base").predict("""Я люблю Spark NLP""")

Model Information

Model Name: roberta_embeddings_ukr_roberta_base
Compatibility: Spark NLP 3.4.2+
License: Open Source
Edition: Official
Input Labels: [sentence, token]
Output Labels: [bert]
Language: uk
Size: 474.2 MB
Case sensitive: true

References

  • https://huggingface.co/youscan/ukr-roberta-base
  • https://dumps.wikimedia.org/ukwiki/latest/ukwiki-latest-pages-articles.xml.bz2
  • https://oscar-public.huma-num.fr/shuffled/uk_dedup.txt.gz
  • https://github.com/youscan/language-models
  • https://twitter.com/vitaliradchenko