1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
|
"""
表格4级向量化器 - 生产级实现
支持HTML/Markdown/Excel多种格式，带缓存和批处理优化
"""
import re
import hashlib
import time
from typing import Optional, List, Dict, Tuple
from dataclasses import dataclass, field
from functools import lru_cache
import logging
logger = logging.getLogger(__name__)
@dataclass
class TableData:
"""表格数据结构"""
table_id: str
table_title: str
headers: List[str]
rows: List[List[str]]
source_format: str
metadata: Dict = field(default_factory=dict)
@property
def row_count(self) -> int:
return len(self.rows)
@property
def col_count(self) -> int:
return len(self.headers)
@property
def cell_count(self) -> int:
return sum(len(row) for row in self.rows)
@dataclass
class VectorizedEntry:
"""向量化条目"""
doc_id: str
vector_level: str
associate_id: str
text_content: str
modal_type: str = "word_table"
business_tag: str = ""
vector: Optional[List[float]] = None
embedding_time: Optional[float] = None
class TableVectorizerPro:
"""
表格4级向量化器（增强版）
改进点：
1. 支持多种表格格式解析
2. 智能空值处理
3. 文本清洗和标准化
4. 批量处理优化
5. 统计和日志记录
"""
def __init__(
self,
max_table_preview_rows: int = 3,
empty_cell_placeholder: str = "-",
enable_text_cleaning: bool = True,
max_text_length: int = 512,
):
self.max_table_preview_rows = max_table_preview_rows
self.empty_cell_placeholder = empty_cell_placeholder
self.enable_text_cleaning = enable_text_cleaning
self.max_text_length = max_text_length
self.stats = {
'tables_processed': 0,
'vectors_generated': {
'table': 0,
'row': 0,
'col': 0,
'cell': 0
},
'processing_time_ms': 0
}
def vectorize(self, table: TableData, business_tag: str = "") -> List[VectorizedEntry]:
"""
对表格执行4级向量化
参数:
table: 表格数据对象
business_tag: 业务标签
返回:
向量化条目列表
"""
start_time = time.time()
logger.info(
f"开始向量化表格: id={table.table_id}, "
f"rows={table.row_count}, cols={table.col_count}"
)
results = []
try:
table_entry = self._vectorize_table_level(table, business_tag)
if table_entry:
results.append(table_entry)
for row_idx, row in enumerate(table.rows):
row_entry = self._vectorize_row_level(
table, row, row_idx, business_tag
)
if row_entry:
results.append(row_entry)
for col_idx in range(table.col_count):
col_entry = self._vectorize_col_level(
table, col_idx, business_tag
)
if col_entry:
results.append(col_entry)
for row_idx, row in enumerate(table.rows):
for col_idx in range(min(len(row), table.col_count)):
cell_entry = self._vectorize_cell_level(
table, row, row_idx, col_idx, business_tag
)
if cell_entry:
results.append(cell_entry)
elapsed_ms = (time.time() - start_time) * 1000
self.stats['tables_processed'] += 1
self.stats['processing_time_ms'] += elapsed_ms
for entry in results:
level = entry.vector_level
if level in self.stats['vectors_generated']:
self.stats['vectors_generated'][level] += 1
logger.info(
f"表格向量化完成: {table.table_id}, "
f"生成 {len(results)} 个向量 (耗时 {elapsed_ms:.1f}ms)"
)
except Exception as e:
logger.error(f"表格向量化异常: {table.table_id}, 错误: {e}")
raise
return results
def _vectorize_table_level(
self,
table: TableData,
business_tag: str
) -> Optional[VectorizedEntry]:
"""
表格级向量化
生成表格的整体摘要
"""
parts = []
if table.table_title:
parts.append(f"表格: {table.table_title}")
headers_str = " | ".join(table.headers)
parts.append(f"表头: {headers_str}")
preview_rows = table.rows[:self.max_table_preview_rows]
for i, row in enumerate(preview_rows):
cells = []
for j, header in enumerate(table.headers):
value = row[j] if j < len(row) else self.empty_cell_placeholder
value = self._clean_text(value)
if value and value != self.empty_cell_placeholder:
cells.append(f"{header}:{value}")
if cells:
parts.append(f"第{i+1}行 " + ", ".join(cells))
if len(table.rows) > self.max_table_preview_rows:
parts.append(f"... 共 {len(table.rows)} 行数据")
text_content = "\n".join(parts)
if not text_content.strip():
return None
return VectorizedEntry(
doc_id=f"{self._generate_doc_id(table.metadata.get('doc_id', ''), 'tbl', table.table_id)}",
vector_level="table",
associate_id=table.table_id,
text_content=text_content[:self.max_text_length],
modal_type="word_table",
business_tag=business_tag
)
def _vectorize_row_level(
self,
table: TableData,
row: List[str],
row_idx: int,
business_tag: str
) -> Optional[VectorizedEntry]:
"""
行级向量化
将一行数据转换为自然语言描述
"""
cells = []
for j, header in enumerate(table.headers):
value = row[j] if j < len(row) else self.empty_cell_placeholder
value = self._clean_text(value)
if value and value != self.empty_cell_placeholder:
cells.append(f"{header}={value}")
if not cells:
return None
text_content = ", ".join(cells)
return VectorizedEntry(
doc_id=self._generate_doc_id(
table.metadata.get('doc_id', ''),
'row',
f"{table.table_id}_{row_idx}"
),
vector_level="row",
associate_id=f"row_{table.table_id}_{row_idx}",
text_content=text_content[:self.max_text_length],
modal_type="word_table",
business_tag=business_tag
)
def _vectorize_col_level(
self,
table: TableData,
col_idx: int,
business_tag: str
) -> Optional[VectorizedEntry]:
"""
列级向量化
将一列数据转换为列表形式
"""
if col_idx >= len(table.headers):
return None
header_name = table.headers[col_idx]
values = []
for row in table.rows:
value = row[col_idx] if col_idx < len(row) else self.empty_cell_placeholder
value = self._clean_text(value)
if value and value != self.empty_cell_placeholder:
values.append(value)
if not values:
return None
text_content = f"{header_name}: " + ", ".join(values)
return VectorizedEntry(
doc_id=self._generate_doc_id(
table.metadata.get('doc_id', ''),
'col',
f"{table.table_id}_{col_idx}"
),
vector_level="col",
associate_id=f"col_{table.table_id}_{col_idx}",
text_content=text_content[:self.max_text_length],
modal_type="word_table",
business_tag=business_tag
)
def _vectorize_cell_level(
self,
table: TableData,
row: List[str],
row_idx: int,
col_idx: int,
business_tag: str
) -> Optional[VectorizedEntry]:
"""
单元格级向量化
精确到单个单元格
"""
if col_idx >= len(row) or col_idx >= len(table.headers):
return None
value = row[col_idx]
value = self._clean_text(value)
if not value or value == self.empty_cell_placeholder or not value.strip():
return None
header_name = table.headers[col_idx] if col_idx < len(table.headers) else ""
if header_name:
text_content = f"{header_name}: {value}"
else:
text_content = value
return VectorizedEntry(
doc_id=self._generate_doc_id(
table.metadata.get('doc_id', ''),
'cell',
f"{table.table_id}_{row_idx}_{col_idx}"
),
vector_level="cell",
associate_id=f"cell_{table.table_id}_{row_idx}_{col_idx}",
text_content=text_content[:self.max_text_length],
modal_type="word_table",
business_tag=business_tag
)
def _clean_text(self, text: str) -> str:
"""
文本清洗
去除多余空白、特殊字符等
"""
if not self.enable_text_cleaning:
return text
if not isinstance(text, str):
text = str(text)
text = text.strip()
text = re.sub(r'\s+', ' ', text)
text = re.sub(r'[^\u4e00-\u9fa5a-zA-Z0-9\s\-.,;:!?%()（）。，；：！？]', '', text)
return text
def _generate_doc_id(
self,
base_doc_id: str,
level: str,
identifier: str
) -> str:
"""
生成唯一的文档ID
格式: {base_doc_id}_{level}_{identifier}
"""
if not base_doc_id:
base_doc_id = "unknown"
safe_identifier = re.sub(r'[^\w\-]', '_', str(identifier))
return f"{base_doc_id}_{level}_{safe_identifier}"
def get_statistics(self) -> Dict:
"""获取处理统计信息"""
return {
**self.stats,
'avg_processing_time_ms': (
self.stats['processing_time_ms'] / self.stats['tables_processed']
if self.stats['tables_processed'] > 0 else 0
),
'avg_vectors_per_table': (
sum(self.stats['vectors_generated'].values()) / self.stats['tables_processed']
if self.stats['tables_processed'] > 0 else 0
)
}
def reset_statistics(self):
"""重置统计信息"""
self.stats = {
'tables_processed': 0,
'vectors_generated': {'table': 0, 'row': 0, 'col': 0, 'cell': 0},
'processing_time_ms': 0
}
|