Edit on GitHub
  • John Snow Labs
  • Nov 30, 2022

Legal Romanian NER (RONEC dataset)

  • ro
  • ner
  • legal
  • ronec
  • licensed

Description

The legner_ronec is a Named Entity Recognition model trained on RONEC (ROmanian Named Entity Corpus). Unlike the original dataset, it has been trained with the following classes:

  • PERSON - proper nouns or pronouns if they refer to a person
  • LOC - location or geo political entity
  • ORG - organization
  • LANGUAGE - language
  • NAT_REL_POL - national, religious or political organizations
  • DATETIME - a time and date in any format, including references to time (e.g. ‘yesterday’)
  • MONEY - a monetary value, numeric or otherwise
  • NUMERIC - a simple numeric value, represented as digits or words
  • ORDINAL - an ordinal value like ‘first’, ‘third’, etc.
  • WORK_OF_ART - a work of art like a named TV show, painting, etc.
  • EVENT - a named recognizable or periodic major event

Predicted Entities

DATETIME, EVENT, LANGUAGE, LOC, MONEY, NAT_REL_POL, NUMERIC, ORDINAL, ORG, PERSON, WORK_OF_ART

Copy S3 URI

How to use

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

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

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

embeddings = nlp.BertEmbeddings.pretrained("bert_base_cased", "ro")\
    .setInputCols("sentence", "token")\
    .setOutputCol("embeddings")\
    .setMaxSentenceLength(512)\
    .setCaseSensitive(True)

ner_model = legal.NerModel.pretrained("legner_ronec", "ro", "legal/models")\
    .setInputCols(["sentence", "token", "embeddings"])\
    .setOutputCol("ner")

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

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

model = pipeline.fit(spark.createDataFrame([[""]]).toDF("text"))

data = spark.createDataFrame([["""Guvernul de stânga italian, condus de premierul Romano Prodi, a devenit după numirea a încă trei secretari de stat, cel mai numeros Executiv din istoria Republicii italiene, având 102 membri."""]]).toDF("text")

result = model.transform(data)

Results

+----------------------+-----------+
|ner_chunk             |label      |
+----------------------+-----------+
|Guvernul              |ORG        |
|italian               |NAT_REL_POL|
|premierul Romano Prodi|PERSON     |
|trei                  |NUMERIC    |
|secretari             |PERSON     |
|Republicii italiene   |LOC        |
|102                   |NUMERIC    |
|membri                |PERSON     |
+----------------------+-----------+

Model Information

Model Name: legner_ronec
Compatibility: Legal NLP 1.0.0+
License: Licensed
Edition: Official
Input Labels: [sentence, token, embeddings]
Output Labels: [ner]
Language: ro
Size: 16.2 MB

References

Dataset is available here.

Benchmarking

label         precision  recall  f1-score  support
DATETIME      0.90       0.90    0.90      1070
EVENT         0.53       0.68    0.59      116
LANGUAGE      0.98       0.95    0.97      44
LOC           0.91       0.90    0.91      1699
MONEY         0.97       0.97    0.97      130
NAT_REL_POL   0.92       0.94    0.93      510
NUMERIC       0.95       0.95    0.95      970
ORDINAL       0.88       0.93    0.90      183
ORG           0.81       0.83    0.82      779
PERSON        0.89       0.91    0.90      2635
WORK_OF_ART   0.73       0.57    0.64      140
micro-avg     0.89       0.90    0.89      8276
macro-avg     0.86       0.87    0.86      8276
weighted-avg  0.89       0.90    0.89      8276