Wineberto labels

Pretrained model on on wine labels only for named entity recognition that uses bert-base-uncased as the base model.

Model description

How to use

You can use this model directly for named entity recognition like so

>>> from transformers import pipeline
>>> ner = pipeline('ner', model='winberto-labels')
>>> tokens = ner("Heitz Cabernet Sauvignon California Napa Valley Napa US")
>>> for t in toks:
>>>    print(f"{t['word']}: {t['entity_group']}: {t['score']:.5}")

heitz: producer: 0.99758
cabernet: wine: 0.92263
sauvignon: wine: 0.92472
california: region: 0.53502
napa valley: subregion: 0.79638
us: country: 0.93675

Training data

The BERT model was trained on 50K wine labels derived from https://www.liv-ex.com/wwd/lwin/ and manually annotated to capture the following tokens

"1": "B-classification",
"2": "B-country",
"3": "B-producer",
"4": "B-region",
"5": "B-subregion",
"6": "B-vintage",
"7": "B-wine"

Training procedure

model_id = 'bert-base-uncased'
arguments = TrainingArguments(
    evaluation_strategy="epoch",
    learning_rate=2e-5,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=8,
    num_train_epochs=5,
    weight_decay=0.01,
)
...
trainer.train()