Tudorx95/NER_Economic_Political

GLiNER Fine-tuned for Political & Economic NER

Fine-tuned version of urchade/gliner_small-v2.1 on a custom politico-economic NER dataset. Trained to recognize 11 entity types.

Entity types

POLITICIAN, POLITICAL_PARTY, POLITICAL_ORG, FINANCIAL_ORG, ECONOMIC_INDICATOR, POLICY, LEGISLATION, MARKET_EVENT, CURRENCY, TRADE_AGREEMENT, GPE

Performance

Test set: 2124 examples. Evaluation mode: ent_type (label match, ignoring exact boundaries).

Global (micro-averaged):

• Precision: 0.6816
• Recall: 0.8992
• F1: 0.7754

Per label:

Label	Precision	Recall	F1
POLITICIAN	0.603	0.902	0.723
POLITICAL_PARTY	0.737	0.964	0.835
POLITICAL_ORG	0.362	0.428	0.392
FINANCIAL_ORG	0.258	0.452	0.329
ECONOMIC_INDICATOR	0.250	0.800	0.381
POLICY	0.000	0.000	0.000
LEGISLATION	0.238	1.000	0.385
MARKET_EVENT	0.146	0.581	0.234
CURRENCY	0.102	0.433	0.166
TRADE_AGREEMENT	0.162	0.429	0.235
GPE	0.825	0.971	0.892

Usage

from gliner import GLiNER

model = GLiNER.from_pretrained("Tudorx95/NER_Economic_Political")
labels = ["POLITICIAN", "POLITICAL_PARTY", "POLITICAL_ORG", "FINANCIAL_ORG",
          "ECONOMIC_INDICATOR", "POLICY", "LEGISLATION", "MARKET_EVENT",
          "CURRENCY", "TRADE_AGREEMENT", "GPE"]

text = "The Federal Reserve raised rates after President Biden signed the new bill."
entities = model.predict_entities(text, labels, threshold=0.5)
for e in entities:
    print(e["text"], "->", e["label"])

Training details

• Base model: urchade/gliner_small-v2.1
• Training examples: 5747
• Validation examples: 1228
• Epochs: 5
• Batch size: 8
• Learning rate: 5e-06