metadata

license: apache-2.0
datasets:
  - empathyai/grocery-ner-dataset
language:
  - en
base_model:
  - gliner-community/gliner_large-v2.5
pipeline_tag: token-classification
library_name: gliner
tags:
  - ner
  - gliner
  - groceries

Grocery Named Entity Recognition Model

A fine-tuned GLiNER model for identifying grocery items and food categories in text.

Model Description

This model is fine-tuned on the grocery-ner-dataset to identify 14 different categories of grocery items including fruits, vegetables, dairy products, and more.

Supported Entity Types

Fruits Vegetables
Lactose, Diary, Eggs, Cheese, Yoghurt
Meat, Fish, Seafood
Frozen, Prepared Meals
Baking, Cooking
Cereals, Grains, Canned, Seeds
Breads
Snacks, Pastries, Treats
Frozen Desserts
Hot Drinks, Chilled Drinks
Alcoholic Drinks
Spices, Sauces
World Foods
Dietary Restrictions, Health, Allergens, Lifestyle

Training Details

Base Model: gliner-community/gliner_medium-v2.5
Training Data: empathyai/grocery-ner-dataset
Batch Size: 8
Learning Rate: 5e-6
Weight Decay: 0.01
Focal Loss Parameters: alpha=0.75, gamma=2
Training Strategy: Linear learning rate with 10% warmup

Usage Example

!pip install gliner

from gliner import GLiNER

# Load model
model = GLiNER.from_pretrained("empathyai/gliner_large-v2.5-groceries")

labels = [
    "Fruits Vegetables",
    "Lactose, Diary, Eggs, Cheese, Yoghurt",
    "Meat, Fish, Seafood",
    "Frozen, Prepared Meals",
    "Baking, Cooking",
    "Cereals, Grains, Canned, Seeds",
    "Breads",
    "Snacks, Pastries, Treats",
    "Frozen Desserts",
    "Hot Drinks, Chilled Drinks",
    "Alcoholic Drinks",
    "Spices, Sauces",
    "World Foods",
    "Dietary Restrictions, Health, Allergens, Lifestyle"
  ]

# Example text
text = "I need to buy milk, bread, and fresh apples"

# Get predictions
predictions = model.predict_entities(text, labels=labels)
print(predictions)

Limitations

Optimized for English language text only
Best performance on grocery shopping and food-related contexts
May not recognize brand names or regional food items not present in training data