Turkish Named Entity Recognition (from winvoker)

Description

Pretrained Named Entity Recognition model, uploaded to Hugging Face, adapted and imported into Spark NLP. bert-base-turkish-cased-ner-tf is a Turkish model orginally trained by winvoker.

Predicted Entities

LOC, PER, ORG

Download Copy S3 URI

How to use

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

sentenceDetector = SentenceDetectorDLModel.pretrained("sentence_detector_dl", "xx")\
       .setInputCols(["document"])\
       .setOutputCol("sentence")

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

tokenClassifier = BertForTokenClassification.pretrained("bert_ner_bert_base_turkish_cased_ner_tf","tr") \
    .setInputCols(["sentence", "token"]) \
    .setOutputCol("ner")

pipeline = Pipeline(stages=[documentAssembler, sentenceDetector, tokenizer, tokenClassifier])

data = spark.createDataFrame([["Spark NLP'yi seviyorum"]]).toDF("text")

result = pipeline.fit(data).transform(data)
val documentAssembler = new DocumentAssembler() 
          .setInputCol("text") 
          .setOutputCol("document")

val sentenceDetector = SentenceDetectorDLModel.pretrained("sentence_detector_dl", "xx")
       .setInputCols(Array("document"))
       .setOutputCol("sentence")

val tokenizer = new Tokenizer() 
    .setInputCols(Array("sentence"))
    .setOutputCol("token")

val tokenClassifier = BertForTokenClassification.pretrained("bert_ner_bert_base_turkish_cased_ner_tf","tr") 
    .setInputCols(Array("sentence", "token")) 
    .setOutputCol("ner")

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

val data = Seq("Spark NLP'yi seviyorum").toDF("text")

val result = pipeline.fit(data).transform(data)
import nlu
nlu.load("tr.ner.bert.cased_base.by_winvoker").predict("""Spark NLP'yi seviyorum""")

Model Information

Model Name: bert_ner_bert_base_turkish_cased_ner_tf
Compatibility: Spark NLP 3.4.2+
License: Open Source
Edition: Official
Input Labels: [document, token]
Output Labels: [ner]
Language: tr
Size: 412.9 MB
Case sensitive: true
Max sentence length: 128

References

  • https://huggingface.co/winvoker/bert-base-turkish-cased-ner-tf
  • https://github.com/stefan-it/turkish-bert/files/4558187/nerdata.txt
  • https://ieeexplore.ieee.org/document/7495744