Spaces:
Sleeping
Sleeping
Upload app.py
Browse files- src/app.py +170 -0
src/app.py
ADDED
|
@@ -0,0 +1,170 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
import transformers
|
| 3 |
+
from transformers import AutoTokenizer, AutoModelForTokenClassification
|
| 4 |
+
import torch
|
| 5 |
+
from html import escape
|
| 6 |
+
import os
|
| 7 |
+
|
| 8 |
+
# 修复权限问题
|
| 9 |
+
os.environ['STREAMLIT_CONFIG_DIR'] = os.getcwd()
|
| 10 |
+
os.environ['STREAMLIT_GATHER_USAGE_STATS'] = 'false'
|
| 11 |
+
# 设置页面
|
| 12 |
+
st.set_page_config(page_title="NER 实体识别", page_icon="🔍", layout="wide")
|
| 13 |
+
|
| 14 |
+
# 标题和描述
|
| 15 |
+
st.title("🔍 命名实体识别 (NER)")
|
| 16 |
+
st.markdown("使用 `dslim/bert-base-NER` 模型识别文本中的实体并分类")
|
| 17 |
+
|
| 18 |
+
# 初始化模型
|
| 19 |
+
@st.cache_resource
|
| 20 |
+
def load_model():
|
| 21 |
+
tokenizer = AutoTokenizer.from_pretrained("dslim/bert-base-NER")
|
| 22 |
+
model = AutoModelForTokenClassification.from_pretrained("dslim/bert-base-NER")
|
| 23 |
+
return tokenizer, model
|
| 24 |
+
|
| 25 |
+
# 加载模型
|
| 26 |
+
with st.spinner("正在加载模型,请稍候..."):
|
| 27 |
+
try:
|
| 28 |
+
tokenizer, model = load_model()
|
| 29 |
+
st.success("模型加载成功!")
|
| 30 |
+
except Exception as e:
|
| 31 |
+
st.error(f"加载模型时出错: {e}")
|
| 32 |
+
st.stop()
|
| 33 |
+
|
| 34 |
+
# 实体类型颜色映射
|
| 35 |
+
entity_colors = {
|
| 36 |
+
'PER': '#FF6B6B', # 人物 - 红色
|
| 37 |
+
'ORG': '#4ECDC4', # 组织 - 青绿色
|
| 38 |
+
'LOC': '#FFD166', # 地点 - 黄色
|
| 39 |
+
'MISC': '#9E6FDC' # 其他 - 紫色
|
| 40 |
+
}
|
| 41 |
+
|
| 42 |
+
# 实体类型描述
|
| 43 |
+
entity_descriptions = {
|
| 44 |
+
'PER': '人物 (Person)',
|
| 45 |
+
'ORG': '组织 (Organization)',
|
| 46 |
+
'LOC': '地点 (Location)',
|
| 47 |
+
'MISC': '其他 (Miscellaneous)'
|
| 48 |
+
}
|
| 49 |
+
|
| 50 |
+
# 显示实体类型说明
|
| 51 |
+
with st.expander("实体类型说明"):
|
| 52 |
+
cols = st.columns(4)
|
| 53 |
+
for i, (key, value) in enumerate(entity_descriptions.items()):
|
| 54 |
+
color = entity_colors.get(key, '#CCCCCC')
|
| 55 |
+
cols[i].markdown(f"<span style='background-color:{color}; padding:5px; border-radius:5px; color:white;'>{key}</span> - {value}", unsafe_allow_html=True)
|
| 56 |
+
|
| 57 |
+
# 输入文本
|
| 58 |
+
default_text = "My name is Clara and I live in Berkeley, California."
|
| 59 |
+
text_input = st.text_area("输入要分析的文本:", value=default_text, height=100)
|
| 60 |
+
|
| 61 |
+
# 处理按钮
|
| 62 |
+
if st.button("识别实体", type="primary"):
|
| 63 |
+
if not text_input.strip():
|
| 64 |
+
st.warning("请输入一些文本进行分析")
|
| 65 |
+
else:
|
| 66 |
+
with st.spinner("正在分析文本..."):
|
| 67 |
+
try:
|
| 68 |
+
# 对输入文本进行编码
|
| 69 |
+
inputs = tokenizer(text_input, return_tensors="pt", truncation=True, max_length=512)
|
| 70 |
+
|
| 71 |
+
# 使用模型进行预测
|
| 72 |
+
outputs = model(**inputs)
|
| 73 |
+
|
| 74 |
+
# 获取预测结果
|
| 75 |
+
predictions = torch.argmax(outputs.logits, dim=-1)[0]
|
| 76 |
+
|
| 77 |
+
# 解码预测结果
|
| 78 |
+
predicted_labels = [model.config.id2label[t.item()] for t in predictions]
|
| 79 |
+
tokens = tokenizer.convert_ids_to_tokens(inputs["input_ids"][0])
|
| 80 |
+
|
| 81 |
+
# 处理输出
|
| 82 |
+
current_entity = None
|
| 83 |
+
current_tokens = []
|
| 84 |
+
results = []
|
| 85 |
+
|
| 86 |
+
for token, label in zip(tokens, predicted_labels):
|
| 87 |
+
# 跳过特殊token
|
| 88 |
+
if token in ['[CLS]', '[SEP]']:
|
| 89 |
+
continue
|
| 90 |
+
|
| 91 |
+
# 处理子词token
|
| 92 |
+
if token.startswith('##'):
|
| 93 |
+
token = token[2:]
|
| 94 |
+
|
| 95 |
+
# 处理实体标签
|
| 96 |
+
if label != 'O':
|
| 97 |
+
entity_type = label.split('-')[-1]
|
| 98 |
+
|
| 99 |
+
if label.startswith('B-'):
|
| 100 |
+
# 如果是新实体的开始
|
| 101 |
+
if current_entity:
|
| 102 |
+
results.append((' '.join(current_tokens), current_entity))
|
| 103 |
+
current_entity = entity_type
|
| 104 |
+
current_tokens = [token]
|
| 105 |
+
else:
|
| 106 |
+
# 继续当前实体
|
| 107 |
+
current_tokens.append(token)
|
| 108 |
+
else:
|
| 109 |
+
# 如果不是实体
|
| 110 |
+
if current_entity:
|
| 111 |
+
results.append((' '.join(current_tokens), current_entity))
|
| 112 |
+
current_entity = None
|
| 113 |
+
current_tokens = []
|
| 114 |
+
|
| 115 |
+
# 添加最后一个实体
|
| 116 |
+
if current_entity:
|
| 117 |
+
results.append((' '.join(current_tokens), current_entity))
|
| 118 |
+
|
| 119 |
+
# 显示结果
|
| 120 |
+
st.subheader("分析结果")
|
| 121 |
+
|
| 122 |
+
# 创建两列布局
|
| 123 |
+
col1, col2 = st.columns([2, 1])
|
| 124 |
+
|
| 125 |
+
with col1:
|
| 126 |
+
st.markdown("**文本中的实体:**")
|
| 127 |
+
|
| 128 |
+
# 高亮显示文本中的实体
|
| 129 |
+
highlighted_text = text_input
|
| 130 |
+
for entity, e_type in results:
|
| 131 |
+
color = entity_colors.get(e_type, '#CCCCCC')
|
| 132 |
+
highlighted_entity = f"<mark style='background-color: {color}; padding: 2px 4px; border-radius: 4px;'>{entity} [{e_type}]</mark>"
|
| 133 |
+
highlighted_text = highlighted_text.replace(entity, highlighted_entity)
|
| 134 |
+
|
| 135 |
+
st.markdown(highlighted_text, unsafe_allow_html=True)
|
| 136 |
+
|
| 137 |
+
with col2:
|
| 138 |
+
st.markdown("**检测到的实体列表:**")
|
| 139 |
+
|
| 140 |
+
if not results:
|
| 141 |
+
st.info("未检测到任何实体")
|
| 142 |
+
else:
|
| 143 |
+
# 按类型分组实体
|
| 144 |
+
entities_by_type = {}
|
| 145 |
+
for entity, e_type in results:
|
| 146 |
+
if e_type not in entities_by_type:
|
| 147 |
+
entities_by_type[e_type] = []
|
| 148 |
+
if entity not in entities_by_type[e_type]:
|
| 149 |
+
entities_by_type[e_type].append(entity)
|
| 150 |
+
|
| 151 |
+
# 显示每种类型的实体
|
| 152 |
+
for e_type, entities in entities_by_type.items():
|
| 153 |
+
color = entity_colors.get(e_type, '#CCCCCC')
|
| 154 |
+
st.markdown(f"<span style='background-color:{color}; padding:2px 6px; border-radius:4px; color:white;'>{e_type}</span>", unsafe_allow_html=True)
|
| 155 |
+
for entity in entities:
|
| 156 |
+
st.markdown(f"- {entity}")
|
| 157 |
+
|
| 158 |
+
# 显示模型信息
|
| 159 |
+
with st.expander("模型信息"):
|
| 160 |
+
st.markdown("""
|
| 161 |
+
**模型:** dslim/bert-base-NER
|
| 162 |
+
**描述:** 基于BERT的命名实体识别模型,能够识别人物(PER)、组织(ORG)、地点(LOC)和其他(MISC)实体。
|
| 163 |
+
""")
|
| 164 |
+
|
| 165 |
+
except Exception as e:
|
| 166 |
+
st.error(f"分析过程中出错: {e}")
|
| 167 |
+
|
| 168 |
+
# 添加页脚
|
| 169 |
+
st.markdown("---")
|
| 170 |
+
st.markdown("使用 🤗 Transformers 和 Streamlit 构建 | 模型: dslim/bert-base-NER")
|