Address Standardization and Correction Model

This model is t5-base fine-tuned to transform incorrect and non-standard addresses into standardized addresses. , primarily trained for US addresses.

How to use the model

from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

model = AutoModelForSeq2SeqLM.from_pretrained("Hnabil/t5-address-standardizer")
tokenizer = AutoTokenizer.from_pretrained("Hnabil/t5-address-standardizer")

inputs = tokenizer(
  "220, soyth rhodeisland aveune, mason city, iowa, 50401, us",
  return_tensors="pt"
)
outputs = model.generate(**inputs, max_length=100)
print(tokenizer.batch_decode(outputs, skip_special_tokens=True))

# ['220, s rhode island ave, mason city, ia, 50401, us']

Training data

The model has been trained on data from openaddresses.io.