Weakly supervised token classification model for German recipe texts based on bert-base-german-cased.

Code available: https://github.com/chefkoch24/weak-ingredient-recognition

Dataset: https://www.kaggle.com/datasets/sterby/german-recipes-dataset

Recognizes the following entities:
'O': 0,
'B-INGREDIENT': 1,
'I-INGREDIENT': 2,
'B-UNIT': 3,
'I-UNIT': 4,
'B-QUANTITY': 5,
'I-QUANTITY': 6

Training:
epochs: 2
optimizer: Adam
learning rate: 2e-5
max length: 512
batch size: 8
recipes: 7801

The model was trained on single Geforce RTX2080 with 11GB GPU

Metrics on test set (weakly supervised):
accuracy_token 0.9965656995773315
f1_token 0.9965656995773315
precision_token 0.9965656995773315
recall_token 0.9965656995773315

Downloads last month
15
Safetensors
Model size
108M params
Tensor type
I64
·
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.