Financial Pipeline (ORG-PER-ROLE-DATE)

Description

This is a pretrained pipeline to extract Companies (ORG), People (PERSON), Job titles (ROLE) and Dates combining different pretrained NER models to improve coverage.

Predicted Entities

ORG, PERSON, ROLE, DATE

Live Demo Open in Colab Download Copy S3 URI

How to use

from johnsnowlabs import *

deid_pipeline = PretrainedPipeline("finpipe_org_per_role_date", "en", "finance/models")

deid_pipeline.annotate("John Smith works as Computer Engineer at Amazon since 2020")

res = deid_pipeline.annotate("John Smith works as Computer Engineer at Amazon since 2020")
for token, ner in zip(res['token'], res['ner']):
    print(f"{token} ({ner})")

Results

John (B-PERSON)
Smith (I-PERSON)
works (O)
as (O)
Computer (B-ROLE)
Engineer (I-ROLE)
at (O)
Amazon (B-ORG)
since (O)
2020 (B-DATE)

Model Information

References

In-house annotations on legal and financial documents, Ontonotes, Conll 2003, Finsec conll, Cuad dataset, 10k filings

Included Models

• DocumentAssembler
• SentenceDetectorDLModel
• TokenizerModel
• BertEmbeddings
• FinanceNerModel
• FinanceBertForTokenClassification
• NerConverter
• NerConverter
• ChunkMergeModel