Description
IMPORTANT: Don’t run this model on the whole financial report. Instead:
- Split by paragraphs. You can use notebook 1 in Finance or Legal as inspiration;
- Use the
finclf_work_experience_itemText Classifier to select only these paragraphs;
This is a md (medium) version of finre_work_experience model, trained with more data and with unidirectional relation extractions, meaning now the direction of the arrow matters: it goes from the source (chunk1) to the target (chunk2).
This model allows you to analyzed present and past job positions of people, extracting relations between PERSON, ORG, ROLE and DATE. This model requires an NER with the mentioned entities, as finner_org_per_role_date and can also be combined with finassertiondl_past_roles to detect if the entities are mentioned to have happened in the PAST or not (although you can also infer that from the relations as had_role_until).
Predicted Entities
has_role, had_role_until, has_role_from, works_for, has_role_in_company
How to use
documentAssembler = DocumentAssembler()\
.setInputCol("text")\
.setOutputCol("document")
sentencizer = SentenceDetectorDLModel\
.pretrained("sentence_detector_dl", "en") \
.setInputCols(["document"])\
.setOutputCol("sentence")
tokenizer = Tokenizer()\
.setInputCols(["sentence"])\
.setOutputCol("token")
bert_embeddings= BertEmbeddings.pretrained("bert_embeddings_sec_bert_base","en")\
.setInputCols(["sentence", "token"])\
.setOutputCol("bert_embeddings")
ner_model = finance.NerModel.pretrained("finner_org_per_role_date", "en", "finance/models")\
.setInputCols(["sentence", "token", "bert_embeddings"])\
.setOutputCol("ner_orgs")
ner_converter = NerConverter()\
.setInputCols(["sentence","token","ner_orgs"])\
.setOutputCol("ner_chunk")
pos = PerceptronModel.pretrained()\
.setInputCols(["sentence", "token"])\
.setOutputCol("pos")
dependency_parser = DependencyParserModel().pretrained("dependency_conllu", "en")\
.setInputCols(["sentence", "pos", "token"])\
.setOutputCol("dependencies")
re_filter = finance.RENerChunksFilter()\
.setInputCols(["ner_chunk", "dependencies"])\
.setOutputCol("re_ner_chunk")\
.setRelationPairs(["PERSON-ROLE", "PERSON-ORG", "ORG-ROLE", "DATE-ROLE"])
reDL = finance.RelationExtractionDLModel()\
.pretrained('finre_work_experience_md','en','finance/models')\
.setInputCols(["re_ner_chunk", "sentence"])\
.setOutputCol("relations")
nlpPipeline = nlp.Pipeline(stages=[
documentAssembler,
sentencizer,
tokenizer,
bert_embeddings,
ner_model,
ner_converter,
pos,
dependency_parser,
re_filter,
reDL])
empty_data = spark.createDataFrame([[""]]).toDF("text")
model = nlpPipeline.fit(empty_data)
text = f"On December 15, 2021, Anirudh Devgan assumed the role of President and Chief Executive Officer of Cadence, replacing Lip-Bu Tan. Prior to his role as Chief Executive Officer, Dr. Devgan served as President of Cadence. Concurrently, Mr. Tan transitioned to the role of Executive Chair"
lmodel = LightPipeline(model)
results = lmodel.fullAnnotate(text)
rel_df = get_relations_df (results)
rel_df = rel_df[rel_df['relation']!='other']
print(rel_df.to_string(index=False))
print()
Results
relation entity1 entity1_begin entity1_end chunk1 entity2 entity2_begin entity2_end chunk2 confidence
has_role_from DATE 3 19 December 15, 2021 ROLE 57 65 President 0.9532135
has_role_from DATE 3 19 December 15, 2021 ROLE 71 93 Chief Executive Officer 0.91833746
has_role PERSON 22 35 Anirudh Devgan ROLE 57 65 President 0.9993814
has_role PERSON 22 35 Anirudh Devgan ROLE 71 93 Chief Executive Officer 0.9889985
works_for PERSON 22 35 Anirudh Devgan ORG 98 104 Cadence 0.9983778
has_role_in_company ROLE 57 65 President ORG 98 104 Cadence 0.9997348
has_role_in_company ROLE 71 93 Chief Executive Officer ORG 98 104 Cadence 0.99845624
has_role ROLE 150 172 Chief Executive Officer PERSON 175 184 Dr. Devgan 0.85268635
has_role_in_company ROLE 150 172 Chief Executive Officer ORG 209 215 Cadence 0.9976404
has_role PERSON 175 184 Dr. Devgan ROLE 196 204 President 0.99899226
works_for PERSON 175 184 Dr. Devgan ORG 209 215 Cadence 0.99876934
has_role_in_company ROLE 196 204 President ORG 209 215 Cadence 0.9997203
has_role PERSON 232 238 Mr. Tan ROLE 268 282 Executive Chair 0.98612714
Model Information
| Model Name: | finre_work_experience_md |
| Compatibility: | Finance NLP 1.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Language: | en |
| Size: | 405.7 MB |
References
Manual annotations on CUAD dataset, 10K filings and Wikidata
Benchmarking
label Recall Precision F1 Support
had_role_until 1.000 1.000 1.000 117
has_role 0.998 0.995 0.997 649
has_role_from 1.000 1.000 1.000 401
has_role_in_company 0.993 0.993 0.993 268
other 0.996 0.996 0.996 235
works_for 0.994 1.000 0.997 330
Avg. 0.997 0.997 0.997 2035
Weighted-Avg. 0.997 0.997 0.997 2035