Zefty/distilbert-ner-email-org

distilbert-ner-email-org is a fine-tuned version of dslim/distilbert-NER on a set of job application emails. The model is fine-tuned specifically to identify the organizations (ORG) entity, thus it CANNOT identify location (LOC), person (PER), and Miscellaneous (MISC), which is available in the original model. This model is fine-tuned specifically to identify the organizations for a personal side-project of mine to extract out companies from job application emails.

How to use

This model can be utilized with the Transformers pipeline for NER.

from transformers import AutoTokenizer, AutoModelForTokenClassification
from transformers import pipeline

tokenizer = AutoTokenizer.from_pretrained("Zefty/distilbert-ner-email-org")
model = AutoModelForTokenClassification.from_pretrained("Zefty/distilbert-ner-email-org")

nlp = pipeline("ner", model=model, tokenizer=tokenizer, aggregation_strategy="first")
example = "Thank you for Applying to Amazon!"

ner_results = nlp(example)
print(ner_results)

Training data

This model was fine-tuned on a set of job application emails. Instead of using the full tokens from the CoNLL-2003 English Dataset, this dataset only includes the ORG token.

Abbreviation	Description
O	Outside of a named entity
B-ORG	Beginning of an organization right after another organization
I-ORG	organization

Eval results

Metric	Score
Loss	0.0898725837469101
Precision	0.7111111111111111
Recall	0.8205128205128205
F1	0.7619047619047619
Accuracy	0.9760986309658876