Edit on GitHub
  • John Snow Labs
  • Apr 25, 2023

Legal NER in Greek Legislations

  • el
  • legal
  • ner
  • licensed
  • legislation

Description

This Legal NER model extracts the following entities from the Greek legislations:

  • FACILITY: Facilities, such as police stations, departments, etc.
  • GPE: Geopolitical Entity; any reference to a geopolitical entity (e.g., country, city, Greek administrative unit, etc.)
  • LEG_REF: Legislation Reference; any reference to Greek or European legislation
  • ORG: Organization; any reference to a public or private organization
  • PER: Any formal name of a person mentioned in the text
  • PUBLIC_DOC: Public Document Reference

Predicted Entities

FACILITY, GPE, LEG_REF, PUBLIC_DOC, PER, ORG

Download Copy S3 URI

How to use

document_assembler = nlp.DocumentAssembler()\
        .setInputCol("text")\
        .setOutputCol("document")

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

embeddings = nlp.BertEmbeddings.pretrained("bert_embeddings_base_el_cased","el")\
        .setInputCols(["document", "token"])\
        .setOutputCol("embeddings")\
        .setMaxSentenceLength(512)\
        .setCaseSensitive(True)

ner_model = legal.NerModel.pretrained("legner_greek_legislation", "el", "legal/models")\
        .setInputCols(["document", "token", "embeddings"])\
        .setOutputCol("ner")

ner_converter = nlp.NerConverter()\
        .setInputCols(["document", "token", "ner"])\
        .setOutputCol("ner_chunk")

nlpPipeline = nlp.Pipeline(stages=[
        document_assembler,
        tokenizer,
        embeddings,
        ner_model,
        ner_converter])

empty_data = spark.createDataFrame([[""]]).toDF("text")

model = nlpPipeline.fit(empty_data)

text_list = ["""3 του άρθρου 5 του ν. 3148/2003, όπως ισχύει, αντικαθίσταται ως εξής""",
                   """1 του άρθρου 1 ασκούνται πλέον από την ΕΥΔΕ/ΕΣΕΑ μέσα σε δύο μήνες από την έναρξη ισχύος του παρόντος Διατάγματος.""",
                   """Ο Πρόεδρος της Επιτροπής και τα τέσσερα μέλη με ισάριθμα αναπληρωματικά εκλέγονται μεταξύ των δημοτών του Δήμου Κυθήρων.""",
                   """Τη με αριθ. 117/Σ.10η/25 Ιουλ 2016 γνωμοδότηση του Ανωτάτου Στρατιωτικού Συμβουλίου."""]

result = model.transform(spark.createDataFrame(pd.DataFrame({"text" : text_list})))

Results

+----------------------------------------+----------+
|chunk                                   |ner_label |
+----------------------------------------+----------+
|ν. 3148/2003                            |LEG_REF   |
|ΕΥΔΕ/ΕΣΕΑ                               |ORG       |
|Δήμου Κυθήρων                           |GPE       |
|αριθ. 117/Σ.10η/25 Ιουλ 2016 γνωμοδότηση|PUBLIC_DOC|
+----------------------------------------+----------+

Model Information

Model Name: legner_greek_legislation
Compatibility: Legal NLP 1.0.0+
License: Licensed
Edition: Official
Input Labels: [sentence, token, embeddings]
Output Labels: [ner]
Language: el
Size: 16.4 MB

References

In-house annotations

Benchmarking

label         precision  recall  f1-score  support 
FACILITY      0.94       0.80    0.86      64      
GPE           0.77       0.83    0.80      136     
LEG_REF       0.94       0.90    0.92      93      
ORG           0.85       0.74    0.79      173     
PER           0.72       0.71    0.71      58      
PUBLIC_DOC    0.76       0.82    0.79      39      
micro-avg     0.83       0.80    0.81      563     
macro-avg     0.83       0.80    0.81      563     
weighted-avg  0.84       0.80    0.82      563