en_spacy_pii_distilbert

Introduction

The en_spacy_pii_distilbert model is designed for token classification tasks, specifically Named Entity Recognition (NER) using the spaCy library. It focuses on identifying Personally Identifiable Information (PII) in English text.

Architecture

• Model: DistilBERT
• Library: spaCy
• Pipeline Components: Transformer and NER
• Label Scheme:
  • DATE_TIME
  • LOC (Location)
  • NRP (Non-personal Names)
  • ORG (Organization)
  • PER (Person)

Training

The model was trained on the beki/privy dataset, which is specifically curated for structured PII detection. The dataset was generated by Privy, a tool for generating PII data. The model achieved the following metrics:

• NER Precision: 95.30%
• NER Recall: 95.54%
• NER F Score: 95.42%
• Transformer Loss: 61154.85
• NER Loss: 56001.88

Guide: Running Locally

To run the en_spacy_pii_distilbert model locally, follow these steps:

1. Install spaCy and Model:

   pip install spacy
   pip install en_spacy_pii_distilbert==0.0.0
   

2. Load the Model:

   import spacy
   nlp = spacy.load("en_spacy_pii_distilbert")
   

3. Run Inference:

   doc = nlp("SELECT shipping FROM users WHERE shipping = '201 Thayer St Providence RI 02912'")
   for ent in doc.ents:
       print(ent.text, ent.label_)
   

For optimal performance, especially on large datasets, consider using cloud GPUs from providers such as AWS, Google Cloud, or Azure.

License

The en_spacy_pii_distilbert model is provided under the MIT License. This allows for broad use and modification, subject to the license terms. The model was authored by Benjamin Kilimnik.