Update README.md

README.md

@@ -1,3 +1,57 @@
 ---
 license: afl-3.0
+language:
+- zh
 ---
+
+# 中文词语分类
+
+本模型对中文词语进行分类（多标签）。对于一个中文词语，其会被分为一个或多个类别，类别有如下：
+
+```
+"1": "人文科学",
+"2": "农林渔畜",
+"3": "医学",
+"4": "城市信息大全",
+"5": "娱乐",
+"6": "工程与应用科学",
+"7": "生活",
+"8": "电子游戏",
+"9": "社会科学",
+"10": "自然科学",
+"11": "艺术",
+"12": "运动休闲"
+```
+
+> 类别来源于[搜狗词汇的类型](https://pinyin.sogou.com/dict/cate/index/167)
+
+# 使用样例
+
+```python
+import torch
+from transformers import AutoTokenizer, BertForSequenceClassification
+
+model_path = "iioSnail/bert-base-chinese-word-classifier"
+
+tokenizer = AutoTokenizer.from_pretrained(model_path)
+model = BertForSequenceClassification.from_pretrained(model_path)
+
+words = ["2型糖尿病", "太古里", "跑跑卡丁车", "河豚"]
+inputs = tokenizer(words, return_tensors='pt', padding=True)
+outputs = model(**inputs).logits
+outputs = outputs.sigmoid()
+preds = outputs > 0.5
+for i, pred in enumerate(preds):
+    pred = torch.argwhere(pred).view(-1)
+    labels = [model.config.id2label[int(id)] for id in pred]
+    print(words[i], ":", labels)
+```
+
+输出：
+
+```
+2型糖尿病 : ['医学']
+太古里 : ['城市信息大全']
+跑跑卡丁车 : ['电子游戏']
+河豚 : ['人文科学', '娱乐', '电子游戏', '自然科学']
+```