zexho/uae_address_roberta_v1.0
110
NER model (uaexmlroberta_base)
地址命名实体识别(NER)模型。
- 任务: Token Classification (NER)
- 国家/区域: 使用阿拉伯语 + 英语的国家地区
- 标签数: 17
使用方式
加载到 transformers 中进行推理：
from transformers import AutoTokenizer, AutoModelForTokenClassification
model_id = "zexho/uae_address_roberta_v1.0"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForTokenClassification.from_pretrained(model_id, ignore_mismatched_sizes=True)
text = "مثال على العنوان"
inputs = tokenizer(text.split(), is_split_into_words=True, return_tensors="pt", truncation=True)
outputs = model(**inputs)注意: 该模型来自自定义训练代码，参数名可能与标准 BertForTokenClassification 对齐；若出现不匹配，可设置 ignore_mismatched_sizes=True 或自行适配。
标签映射
"id2label": {
"0": "O",
"1": "B-COUNTRY",
"2": "I-COUNTRY",
"3": "B-EMIRATE",
"4": "I-EMIRATE",
"5": "B-CITY",
"6": "I-CITY",
"7": "B-SUB_AREA",
"8": "I-SUB_AREA",
"9": "B-COMPOUND",
"10": "I-COMPOUND",
"11": "B-STREET",
"12": "I-STREET",
"13": "B-BUILDING",
"14": "I-BUILDING",
"15": "B-HOUSE_NUMBER",
"16": "I-HOUSE_NUMBER"
}评估效果