Merge pull request #44 from pekopoke/dev

add extractor version in results 
fix formula match in table and dollar

Author: pekopoke

tests/test_formula_extraction.py

@@ -167,5 +167,19 @@ def test_formula_within_table(self):
         self.assertIn('| 公式类型 | 示例 |', result['table'])
 
 
+    def test_dollar_within_table(self):
+        """测试表格中的转义$包裹的内容不要被提取"""
+
+        text = """
+        <table><tbody><tr><td><table><tbody><tr><td><table><tbody><tr><td><strong>Better Management of /$800 Bln Forex Reserves Urged</strong></td></tr></tbody></table></td></tr><tr><td><p>A number of political advisors on Sunday called for more rationally managing China's massive foreign exchange reserves, which doubled over the 2004-05 period to an equivalent of US /$818.9 billion, second only to Japan.</p><p>The quick buildup is largely a result of China's booming exports and foreign exchange controls by the government, as well as speculation on the yuan's rise, industry watchers agree.</p><p>A big part of China's foreign exchange reserves are US dollar-denominated assets, including bonds issued by the US government. "Risks in the international foreign exchange market should be lowered when China manages its reserves," said Professor Guo Guoqing of a business school of the People's University of China.</p><p>Guo, a member of the National Committee of the Chinese People's Political Consultative Conference (CPPCC), the country's top advisory body, urged the government to cut back on subsidies for exports and take other measures to reduce foreign trade surpluses appropriately and achieve the balance in international payments.</p><p>Part of the reserves should be channeled into the imports of more high-tech machinery, equipment and other products, he suggested on the sidelines of the CPPCC's annual session.</p><p>The United States has been contending that the value of yuan, also known as renminbi or RMB, is too low, giving Chinese exporters an "unfair" advantage. But China said its huge trade surpluses are also a result of the US reluctance to export goods involving state-of-the-art technologies.</p><p>Fu Rui, also a CPPCC member, said with ample foreign exchange reserves, China could intentionally bulk up the reserves of strategic resources.</p><p>The international consensus is a country's rational foreign exchange reserves should equal to its imports demand for a full quarter. Also taking into consideration of payments for foreign debts, returns for foreign investors and other demands in China, many believe it is enough for the country to retain US/$300 billion.</p><p>But Lin Yifu, a popular economist, underscored China's per capita foreign exchange reserves remains not large - less than one-tenth of Japan's and far below that of Hong Kong and Singapore.</p><p>The reserves were "tremendous fruits" from China's reform and opening-up drive, he said.</p><p>His remarks were echoed by Xiao Zhuoji, a well-known economics professor with Beijing University. "The rise of foreign exchange reserves reflects China's fast, sustained economic growth and sound international payments," he said.</p><p>"The reserves are of significant importance to upgrade the China image in the international economic arena, strengthen the nation's macro-control capabilities and guard against financial risks," added Xiao, a Standing Committee member of the CPPCC National Committee.</p><p>But as the People's Bank of China, or the central bank, has to buy foreign exchange reserves under the current foreign exchange control policies, the country's monetary base will be enlarged, increasing its inflationary pressure and difficulties on macro-economic controls, analysts acknowledge.</p><p>Another prevailing view is that China's hefty foreign exchange reserves actually "occupied" large amounts of fund resources that otherwise can be diverted for domestic investment and consumption.</p><p>Some CPPCC members said they believe it is already "meaningless" now to talk about whether China's foreign exchange reserves size is big or not. "The key lies on how to raise the reserves' yields."</p><p>"If the annual yields from foreign exchange reserves could reach a stable 5 percent, the nation will reap in 300 billion yuan a year. What a big fortune!" one advisor told Xinhua.</p><p>Central banker Zhou Xiaochuan reiterated earlier that China will "pay attention to and maintain the flexibility" of foreign reserves structure, which is unknown to the public.</p></td></tr></tbody></table></td></tr></tbody></table>
+        """
+
+        result = self.metric._extract_from_markdown(text)
+
+        # 验证表格中的转义$包裹的内容不要被提取
+        self.assertNotIn('800', result['formula'])
+
+
+
 if __name__ == '__main__':
     unittest.main()

webmainbench/data/saver.py

@@ -126,32 +126,45 @@ def save_summary_report(results: Union["EvaluationResult", List["EvaluationResul
             file_path: Output CSV file path
         """
         import csv
-        
+        from importlib import metadata as importlib_metadata
+
         file_path = Path(file_path)
         file_path.parent.mkdir(parents=True, exist_ok=True)
-        
-        # Convert EvaluationResult objects to dicts and ensure we have a list
+
+        # 转换结果为字典列表
         def to_dict_if_needed(item):
             return item.to_dict() if hasattr(item, 'to_dict') else item
-        
+
         if isinstance(results, list):
             results_list = [to_dict_if_needed(item) for item in results]
         else:
             results_list = [to_dict_if_needed(results)]
-        
-        # Prepare CSV data
+
         csv_data = []
-        
         for result in results_list:
-            # Extract basic info
             metadata = result.get('metadata', {})
             error_analysis = result.get('error_analysis', {})
-            
+
+            # 获取抽取器版本
+            extractor_name = metadata.get('extractor_name', 'unknown')
+            try:
+                # 映射抽取器名称到包名
+                package_mapping = {
+                    'llm-webkit': 'llm_web_kit',
+                    'magic-html': 'magic_html',
+                    'trafilatura': 'trafilatura',
+                    'resiliparse': 'resiliparse'
+                }
+                package_name = package_mapping.get(extractor_name, extractor_name)
+                extractor_version = importlib_metadata.version(package_name)
+            except importlib_metadata.PackageNotFoundError:
+                extractor_version = 'unknown'
             row = {
                 'extractor': metadata.get('extractor_name', 'unknown'),
                 'dataset': metadata.get('dataset_name', 'unknown'),
                 'total_samples': metadata.get('total_samples', 0),
-                'success_rate': error_analysis.get('success_rate', 0.0)
+                'success_rate': error_analysis.get('success_rate', 0.0),
+                 'extractor_version': extractor_version,
             }
 
             # Add all available metrics from overall_metrics
@@ -170,7 +183,7 @@ def get_sort_key(row):
         # Write CSV file
         if csv_data:
             # Define field order: basic info first, then overall, then other metrics alphabetically
-            basic_fields = ['extractor', 'dataset', 'total_samples', 'success_rate']
+            basic_fields = ['extractor','extractor_version', 'dataset', 'total_samples', 'success_rate']
 
             # Get all metric fields from the data
             all_fields = set()

webmainbench/extractors/llm_webkit_extractor.py

@@ -155,7 +155,7 @@ def process_logit(self, prompt_token_ids: List[int], input_ids: List[int], logit
 class LlmWebkitExtractor(BaseExtractor):
     """Advanced LLM-WebKit extractor with intelligent content classification."""
 
-    version = "2.0.0"
+    version = "4.0.1"
     description = "Advanced LLM-WebKit extractor with intelligent content classification"
 
     # 分类提示模板

webmainbench/metrics/base.py

@@ -307,7 +307,8 @@ def _extract_from_markdown(text: str) -> Dict[str, str]:
             # r'(?<!\\)\$([^$\n\w][^$\n]*[^$\n\w])\$(?![\\\$])',
             r'(?<!\\)\$\$(.*?)(?<!\\)\$\$',  # 行间 $$...$$，确保 $ 没有被转义
             r'(?<!\\)\\\[(.*?)(?<!\\)\\\]',  # 行间 \[...\]，确保 \ 没有被转义
-            r'(?<!\\)\$(.*?)(?<!\\)\$',  # 行内 $...$，确保 $ 没有被转义
+            # r'(?<!\\)\$(.*?)(?<!\\)\$',  # 行内 $...$，确保 $ 没有被转义
+            r'(?<!\\)\$(.*?)(?<!\\)\$(?!\d)',  # 第二个$后面不能是数字
             r'(?<!\\)\\\((.*?)(?<!\\)\\\)',  # 行内 \(...\)，确保 \ 没有被转义
         ]