PII NER Azerbaijani is a fine-tuned Named Entity Recognition (NER) model based on XLM-RoBERTa. It is trained on Azerbaijani pii data for classification personally identifiable information such as names, dates of birth, cities, addresses, and phone numbers from text.

Model Details

Base Model: XLM-RoBERTa
Training Metrics:
- Epoch 1: Training Loss: 0.156, Validation Loss: 0.1309, Precision: 0.7794, Recall: 0.7940, F1: 0.7866, Accuracy: 0.9590
- Epoch 2: Training Loss: 0.1196, Validation Loss: 0.1172, Precision: 0.8042, Recall: 0.8078, F1: 0.8060, Accuracy: 0.9618
- Epoch 3: Training Loss: 0.1069, Validation Loss: 0.1129, Precision: 0.8096, Recall: 0.8213, F1: 0.8154, Accuracy: 0.9639
Test Metrics:
- Loss: 0.11616, Precision: 0.80187, Recall: 0.80821, F1: 0.80503, Accuracy: 0.96264

Entities (id2label)

python

{
"O",
"VEHICLEVRM",
"HEIGHT",
"USERNAME",
"FIRSTNAME",
"BUILDINGNUMBER",
"SEX",
"PHONENUMBER",
"CURRENCY",
"CREDITCARDISSUER",
"CURRENCYNAME",
"MAC",
"MIDDLENAME",
"TIME",
"EYECOLOR",
"CURRENCYSYMBOL",
"GENDER",
"URL",
"CURRENCYCODE",
"ZIPCODE",
"CREDITCARDCVV",
"JOBTITLE",
"PHONEIMEI",
"COUNTY",
"JOBTYPE",
"LITECOINADDRESS",
"COMPANYNAME",
"ORDINALDIRECTION",
"MASKEDNUMBER",
"USERAGENT",
"LASTNAME",
"SSN",
"STREET",
"SECONDARYADDRESS",
"STATE",
"ETHEREUMADDRESS",
"AMOUNT",
"ACCOUNTNUMBER",
"CITY",
"CREDITCARDNUMBER",
"BIC",
"EMAIL",
"NEARBYGPSCOORDINATE",
"PIN",
"ACCOUNTNAME",
"VEHICLEVIN",
"PREFIX",
"JOBAREA",
"AGE",
"PASSWORD",
"DOB",
"BITCOINADDRESS",
"IBAN",
"IP",
"DATE"
}

Usage

To use the model for spell correction:

python

import torch
from transformers import AutoTokenizer, AutoModelForTokenClassification

model_id = "LocalDoc/private_ner_azerbaijani"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForTokenClassification.from_pretrained(model_id)

test_text = (
    "Salam, mənim adım Əli Hüseynovdur. Doğum tarixim 15.05.1990-dır. Bakı şəhərində, Nizami küçəsində, 25/31 ünvanında yaşayıram. Telefon nömrəm +994552345678-dir."
)

inputs = tokenizer(test_text, return_tensors="pt", return_offsets_mapping=True)

offset_mapping = inputs.pop("offset_mapping")

with torch.no_grad():
    outputs = model(**inputs)

predictions = torch.argmax(outputs.logits, dim=2)

tokens = tokenizer.convert_ids_to_tokens(inputs["input_ids"][0])
offset_mapping = offset_mapping[0].tolist()
predicted_labels = [model.config.id2label[pred.item()] for pred in predictions[0]]
word_ids = inputs.word_ids(batch_index=0)

aggregated = []
prev_word_id = None
for idx, word_id in enumerate(word_ids):
    if word_id is None:
        continue
    if word_id != prev_word_id:
        aggregated.append({
            "word_id": word_id,
            "tokens": [tokens[idx]],
            "offsets": [offset_mapping[idx]],
            "label": predicted_labels[idx]
        })
    else:
        aggregated[-1]["tokens"].append(tokens[idx])
        aggregated[-1]["offsets"].append(offset_mapping[idx])
    prev_word_id = word_id

entities = []
current_entity = None
for word in aggregated:
    if word["label"] == "O":
        if current_entity is not None:
            entities.append(current_entity)
            current_entity = None
    else:
        if current_entity is None:
            current_entity = {
                "type": word["label"],
                "start": word["offsets"][0][0],
                "end": word["offsets"][-1][1]
            }
        else:
            if word["label"] == current_entity["type"]:
                current_entity["end"] = word["offsets"][-1][1]
            else:
                entities.append(current_entity)
                current_entity = {
                    "type": word["label"],
                    "start": word["offsets"][0][0],
                    "end": word["offsets"][-1][1]
                }
if current_entity is not None:
    entities.append(current_entity)

for entity in entities:
    entity["text"] = test_text[entity["start"]:entity["end"]]

for entity in entities:
    print(entity)

json

{'type': 'FIRSTNAME', 'start': 18, 'end': 21, 'text': 'Əli'}
{'type': 'LASTNAME', 'start': 22, 'end': 34, 'text': 'Hüseynovdur.'}
{'type': 'DOB', 'start': 49, 'end': 64, 'text': '15.05.1990-dır.'}
{'type': 'STREET', 'start': 81, 'end': 87, 'text': 'Nizami'}
{'type': 'BUILDINGNUMBER', 'start': 99, 'end': 104, 'text': '25/31'}
{'type': 'PHONENUMBER', 'start': 141, 'end': 159, 'text': '+994552345678-dir.'}

License

This model licensed under the CC BY-NC-ND 4.0 license. What does this license allow?

Attribution: You must give appropriate credit, provide a link to the license, and indicate if changes were made.
Non-Commercial: You may not use the material for commercial purposes.
No Derivatives: If you remix, transform, or build upon the material, you may not distribute the modified material.

For more information, please refer to the CC BY-NC-ND 4.0 license.

Contact

For more information, questions, or issues, please contact LocalDoc at [v.resad.89@gmail.com].