Description
The dataset consists of 12 documents taken from EUR-Lex, a multilingual corpus of court decisions and legal dispositions in the 24 official languages of the European Union.
This model extracts ADDRESS, AMOUNT, DATE, ORGANISATION, and PERSON entities from French documents.
Predicted Entities
ADDRESS, AMOUNT, DATE, ORGANISATION, PERSON
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_embeddings_base_fr_cased", "fr")\
.setInputCols(["sentence", "token"])\
.setOutputCol("embeddings")\
.setMaxSentenceLength(512)\
.setCaseSensitive(True)
ner_model = legal.NerModel.pretrained("legner_mapa", "fr", "legal/models")\
.setInputCols(["sentence", "token", "embeddings"])\
.setOutputCol("ner")
ner_converter = nlp.NerConverter()\
.setInputCols(["sentence", "token", "ner"])\
.setOutputCol("ner_chunk")
nlpPipeline = nlp.Pipeline(stages=[
document_assembler,
sentence_detector,
tokenizer,
embeddings,
ner_model,
ner_converter])
empty_data = spark.createDataFrame([[""]]).toDF("text")
model = nlpPipeline.fit(empty_data)
text = ["""Heeren, administrateur, vu la phase écrite de la procédure et à la suite de l’audience du 28 novembre 2017, rend le présent Arrêt Antécédents du litige 1 La requérante, Foshan Lihua Ceramic Co."""]
result = model.transform(spark.createDataFrame([text]).toDF("text"))
Results
+-----------------------+------------+
|chunk |ner_label |
+-----------------------+------------+
|Heeren |PERSON |
|28 novembre 2017 |DATE |
|Foshan Lihua Ceramic Co|ORGANISATION|
+-----------------------+------------+
Model Information
| Model Name: | legner_mapa |
| Compatibility: | Legal NLP 1.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Input Labels: | [sentence, token, embeddings] |
| Output Labels: | [ner] |
| Language: | fr |
| Size: | 1.4 MB |
References
The dataset is available here.
Benchmarking
label precision recall f1-score support
ADDRESS 1.00 1.00 1.00 11
AMOUNT 1.00 1.00 1.00 4
DATE 1.00 0.96 0.98 28
ORGANISATION 1.00 0.95 0.98 22
PERSON 0.94 0.94 0.94 31
macro-avg 0.98 0.96 0.97 96
macro-avg 0.99 0.97 0.98 96
weighted-avg 0.98 0.96 0.97 96