| { | |
| "model_type": "CRF", | |
| "task": "token-classification", | |
| "serialization": "cloudpickle", | |
| "framework": "sklearn-crfsuite", | |
| "language": "pt-BR", | |
| "dataset": "Brazilian Address NER", | |
| "version": "2.0.0", | |
| "architecture": { | |
| "algorithm": "lbfgs", | |
| "transition_features": "all_possible", | |
| "pipeline": [ | |
| "AddressFeatureExtractor", | |
| "CRFTagger" | |
| ] | |
| }, | |
| "hyperparameters": { | |
| "algorithm": "lbfgs", | |
| "c1": 0.1, | |
| "c2": 0.1, | |
| "max_iterations": 100, | |
| "all_possible_transitions": true | |
| }, | |
| "labels": [ | |
| "B-bairro", | |
| "B-cep", | |
| "B-complemento", | |
| "B-estado", | |
| "B-logradouro", | |
| "B-municipio", | |
| "B-numero", | |
| "B-tipo_logradouro", | |
| "I-bairro", | |
| "I-complemento", | |
| "I-logradouro", | |
| "I-municipio", | |
| "I-tipo_logradouro" | |
| ], | |
| "num_labels": 13, | |
| "id2label": { | |
| "0": "B-bairro", | |
| "1": "B-cep", | |
| "2": "B-complemento", | |
| "3": "B-estado", | |
| "4": "B-logradouro", | |
| "5": "B-municipio", | |
| "6": "B-numero", | |
| "7": "B-tipo_logradouro", | |
| "8": "I-bairro", | |
| "9": "I-complemento", | |
| "10": "I-logradouro", | |
| "11": "I-municipio", | |
| "12": "I-tipo_logradouro" | |
| }, | |
| "label2id": { | |
| "B-bairro": 0, | |
| "B-cep": 1, | |
| "B-complemento": 2, | |
| "B-estado": 3, | |
| "B-logradouro": 4, | |
| "B-municipio": 5, | |
| "B-numero": 6, | |
| "B-tipo_logradouro": 7, | |
| "I-bairro": 8, | |
| "I-complemento": 9, | |
| "I-logradouro": 10, | |
| "I-municipio": 11, | |
| "I-tipo_logradouro": 12 | |
| }, | |
| "training": { | |
| "accuracy": 0.999468273661822, | |
| "trained_at": "2025-11-19T15:36:55.624329", | |
| "seed": 42 | |
| }, | |
| "feature_extraction": { | |
| "method": "morphological + contextual", | |
| "morphological_features": [ | |
| "word_lowercase", | |
| "is_digit", | |
| "is_state_abbrev", | |
| "is_title", | |
| "is_upper", | |
| "has_dash", | |
| "is_cep_format", | |
| "is_cep_no_dash", | |
| "word_length", | |
| "common_address_words", | |
| "complement_words" | |
| ], | |
| "contextual_features": [ | |
| "previous_word", | |
| "next_word", | |
| "BOS_marker", | |
| "EOS_marker" | |
| ], | |
| "note": "CRF uses same morphological features as HMM plus contextual features" | |
| } | |
| } |