style(sentiment_analyzer): format file

ghmark675 · 666ghj
Commit aa11c529c8707615a2f40d8a301cd740f345dbad aa11c529 1 parent 71f4b3ad
Showing 1 changed file with 112 additions and 48 deletions
InsightEngine/tools/sentiment_analyzer.py
--- a/InsightEngine/tools/sentiment_analyzer.py
View file @aa11c52
+++ b/InsightEngine/tools/sentiment_analyzer.py
View file @aa11c52
@@ -11,6 +11,7 @@ import re
 try:
     import torch
+
     TORCH_AVAILABLE = True
 except ImportError:
     torch = None  # type: ignore
@@ -18,6 +19,7 @@ except ImportError:
 try:
     from transformers import AutoTokenizer, AutoModelForSequenceClassification
+
     TRANSFORMERS_AVAILABLE = True
 except ImportError:
     AutoTokenizer = None  # type: ignore
@@ -28,6 +30,7 @@ except ImportError:
 # INFO：若想跳过情感分析，可手动切换此开关为False
 SENTIMENT_ANALYSIS_ENABLED = True
+
 def _describe_missing_dependencies() -> str:
     missing = []
     if not TORCH_AVAILABLE:
@@ -36,14 +39,21 @@ def _describe_missing_dependencies() -> str:
         missing.append("Transformers")
     return " / ".join(missing)
+
 # 添加项目根目录到路径，以便导入WeiboMultilingualSentiment
-project_root = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-weibo_sentiment_path = os.path.join(project_root, "SentimentAnalysisModel", "WeiboMultilingualSentiment")
+project_root = os.path.dirname(
+    os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+)
+weibo_sentiment_path = os.path.join(
+    project_root, "SentimentAnalysisModel", "WeiboMultilingualSentiment"
+)
 sys.path.append(weibo_sentiment_path)
+
 @dataclass
 class SentimentResult:
     """情感分析结果数据类"""
+
     text: str
     sentiment_label: str
     confidence: float
@@ -56,6 +66,7 @@ class SentimentResult:
 @dataclass
 class BatchSentimentResult:
     """批量情感分析结果数据类"""
+
     results: List[SentimentResult]
     total_processed: int
     success_count: int
@@ -85,7 +96,7 @@ class WeiboMultilingualSentimentAnalyzer:
             1: "负面",
             2: "中性",
             3: "正面",
-            4: "非常正面"
+            4: "非常正面",
         }
         if not SENTIMENT_ANALYSIS_ENABLED:
@@ -96,9 +107,13 @@ class WeiboMultilingualSentimentAnalyzer:
         if self.is_disabled:
             reason = self.disable_reason or "Sentiment analysis disabled."
-            print(f"WeiboMultilingualSentimentAnalyzer initialized but disabled: {reason}")
+            print(
+                f"WeiboMultilingualSentimentAnalyzer initialized but disabled: {reason}"
+            )
         else:
-            print("WeiboMultilingualSentimentAnalyzer 已创建，调用 initialize() 来加载模型")
+            print(
+                "WeiboMultilingualSentimentAnalyzer 已创建，调用 initialize() 来加载模型"
+            )
     def disable(self, reason: Optional[str] = None, drop_state: bool = False) -> None:
         """Disable sentiment analysis, optionally clearing loaded resources."""
@@ -130,7 +145,11 @@ class WeiboMultilingualSentimentAnalyzer:
         if torch.cuda.is_available():
             return torch.device("cuda")
         mps_backend = getattr(torch.backends, "mps", None)
-        if mps_backend and getattr(mps_backend, "is_available", lambda: False)() and getattr(mps_backend, "is_built", lambda: False)():
+        if (
+            mps_backend
+            and getattr(mps_backend, "is_available", lambda: False)()
+            and getattr(mps_backend, "is_built", lambda: False)()
+        ):
             return torch.device("mps")
         return torch.device("cpu")
@@ -167,12 +186,16 @@ class WeiboMultilingualSentimentAnalyzer:
             if os.path.exists(local_model_path):
                 print("从本地加载模型...")
                 self.tokenizer = AutoTokenizer.from_pretrained(local_model_path)
-                self.model = AutoModelForSequenceClassification.from_pretrained(local_model_path)
+                self.model = AutoModelForSequenceClassification.from_pretrained(
+                    local_model_path
+                )
             else:
                 print("首次使用，正在下载模型到本地...")
                 # 下载并保存到本地
                 self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-                self.model = AutoModelForSequenceClassification.from_pretrained(model_name)
+                self.model = AutoModelForSequenceClassification.from_pretrained(
+                    model_name
+                )
                 # 保存到本地
                 os.makedirs(local_model_path, exist_ok=True)
@@ -227,7 +250,7 @@ class WeiboMultilingualSentimentAnalyzer:
             return ""
         # 去除多余空格
-        text = re.sub(r'\s+', ' ', text.strip())
+        text = re.sub(r"\s+", " ", text.strip())
         return text
@@ -249,7 +272,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 probability_distribution={},
                 success=False,
                 error_message=self.disable_reason or "情感分析功能已禁用",
-                analysis_performed=False
+                analysis_performed=False,
             )
         if not self.is_initialized:
@@ -260,7 +283,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 probability_distribution={},
                 success=False,
                 error_message="模型未初始化，请先调用initialize() 方法",
-                analysis_performed=False
+                analysis_performed=False,
             )
         try:
@@ -275,7 +298,7 @@ class WeiboMultilingualSentimentAnalyzer:
                     probability_distribution={},
                     success=False,
                     error_message="输入文本为空或无效内容",
-                    analysis_performed=False
+                    analysis_performed=False,
                 )
             # 分词编码
@@ -284,7 +307,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 max_length=512,
                 padding=True,
                 truncation=True,
-                return_tensors='pt'
+                return_tensors="pt",
             )
             # 转移到设备
@@ -311,7 +334,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 sentiment_label=label,
                 confidence=confidence,
                 probability_distribution=prob_dist,
-                success=True
+                success=True,
             )
         except Exception as e:
@@ -322,10 +345,12 @@ class WeiboMultilingualSentimentAnalyzer:
                 probability_distribution={},
                 success=False,
                 error_message=f"预测时发生错误: {str(e)}",
-                analysis_performed=False
+                analysis_performed=False,
             )
-    def analyze_batch(self, texts: List[str], show_progress: bool = True) -> BatchSentimentResult:
+    def analyze_batch(
+        self, texts: List[str], show_progress: bool = True
+    ) -> BatchSentimentResult:
         """
         批量情感分析
@@ -343,7 +368,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 success_count=0,
                 failed_count=0,
                 average_confidence=0.0,
-                analysis_performed=not self.is_disabled and self.is_initialized
+                analysis_performed=not self.is_disabled and self.is_initialized,
             )
         if self.is_disabled or not self.is_initialized:
@@ -355,7 +380,7 @@ class WeiboMultilingualSentimentAnalyzer:
                     probability_distribution={},
                     success=False,
                     error_message=self.disable_reason or "情感分析功能不可用",
-                    analysis_performed=False
+                    analysis_performed=False,
                 )
                 for text in texts
             ]
@@ -365,7 +390,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 success_count=0,
                 failed_count=len(texts),
                 average_confidence=0.0,
-                analysis_performed=False
+                analysis_performed=False,
             )
         results = []
@@ -374,7 +399,7 @@ class WeiboMultilingualSentimentAnalyzer:
         for i, text in enumerate(texts):
             if show_progress and len(texts) > 1:
-                print(f"处理进度: {i+1}/{len(texts)}")
+                print(f"处理进度: {i + 1}/{len(texts)}")
             result = self.analyze_single_text(text)
             results.append(result)
@@ -383,7 +408,9 @@ class WeiboMultilingualSentimentAnalyzer:
                 success_count += 1
                 total_confidence += result.confidence
-        average_confidence = total_confidence / success_count if success_count > 0 else 0.0
+        average_confidence = (
+            total_confidence / success_count if success_count > 0 else 0.0
+        )
         failed_count = len(texts) - success_count
         return BatchSentimentResult(
@@ -392,7 +419,7 @@ class WeiboMultilingualSentimentAnalyzer:
             success_count=success_count,
             failed_count=failed_count,
             average_confidence=average_confidence,
-            analysis_performed=True
+            analysis_performed=True,
         )
     def _build_passthrough_analysis(
@@ -400,7 +427,7 @@ class WeiboMultilingualSentimentAnalyzer:
         original_data: List[Dict[str, Any]],
         reason: str,
         texts: Optional[List[str]] = None,
-        results: Optional[List[SentimentResult]] = None
+        results: Optional[List[SentimentResult]] = None,
     ) -> Dict[str, Any]:
         """
         构建在情感分析不可用时的透传结果
@@ -416,7 +443,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 "sentiment_distribution": {},
                 "high_confidence_results": [],
                 "summary": f"情感分析未执行：{reason}",
-                "original_texts": original_data
+                "original_texts": original_data,
             }
         }
@@ -431,9 +458,12 @@ class WeiboMultilingualSentimentAnalyzer:
         return response
-    def analyze_query_results(self, query_results: List[Dict[str, Any]], 
+    def analyze_query_results(
+        self,
+        query_results: List[Dict[str, Any]],
         text_field: str = "content",
-                            min_confidence: float = 0.5) -> Dict[str, Any]:
+        min_confidence: float = 0.5,
+    ) -> Dict[str, Any]:
         """
         对查询结果进行情感分析
         专门用于分析从MediaCrawlerDB返回的查询结果
@@ -452,7 +482,7 @@ class WeiboMultilingualSentimentAnalyzer:
                     "total_analyzed": 0,
                     "sentiment_distribution": {},
                     "high_confidence_results": [],
-                    "summary": "没有内容需要分析"
+                    "summary": "没有内容需要分析",
                 }
             }
@@ -478,7 +508,7 @@ class WeiboMultilingualSentimentAnalyzer:
                     "total_analyzed": 0,
                     "sentiment_distribution": {},
                     "high_confidence_results": [],
-                    "summary": "查询结果中没有找到可分析的文本内容"
+                    "summary": "查询结果中没有找到可分析的文本内容",
                 }
             }
@@ -486,7 +516,7 @@ class WeiboMultilingualSentimentAnalyzer:
             return self._build_passthrough_analysis(
                 original_data=original_data,
                 reason=self.disable_reason or "情感分析模型不可用",
-                texts=texts_to_analyze
+                texts=texts_to_analyze,
             )
         # 执行批量情感分析
@@ -496,14 +526,17 @@ class WeiboMultilingualSentimentAnalyzer:
         if not batch_result.analysis_performed:
             reason = self.disable_reason or "情感分析功能不可用"
             if batch_result.results:
-                candidate_error = next((r.error_message for r in batch_result.results if r.error_message), None)
+                candidate_error = next(
+                    (r.error_message for r in batch_result.results if r.error_message),
+                    None,
+                )
                 if candidate_error:
                     reason = candidate_error
             return self._build_passthrough_analysis(
                 original_data=original_data,
                 reason=reason,
                 texts=texts_to_analyze,
-                results=batch_result.results
+                results=batch_result.results,
             )
         # 统计情感分布
@@ -520,18 +553,22 @@ class WeiboMultilingualSentimentAnalyzer:
                 # 收集高置信度结果
                 if result.confidence >= min_confidence:
-                    high_confidence_results.append({
+                    high_confidence_results.append(
+                        {
                             "original_data": original_item,
                             "sentiment": result.sentiment_label,
                             "confidence": result.confidence,
-                        "text_preview": result.text[:100] + "..." if len(result.text) > 100 else result.text
-                    })
+                            "text_preview": result.text[:100] + "..."
+                            if len(result.text) > 100
+                            else result.text,
+                        }
+                    )
         # 生成情感分析摘要
         total_analyzed = batch_result.success_count
         if total_analyzed > 0:
             dominant_sentiment = max(sentiment_distribution.items(), key=lambda x: x[1])
-            sentiment_summary = f"共分析{total_analyzed}条内容，主要情感倾向为'{dominant_sentiment[0]}'({dominant_sentiment[1]}条，占{dominant_sentiment[1]/total_analyzed*100:.1f}%)"
+            sentiment_summary = f"共分析{total_analyzed}条内容，主要情感倾向为'{dominant_sentiment[0]}'({dominant_sentiment[1]}条，占{dominant_sentiment[1] / total_analyzed * 100:.1f}%)"
         else:
             sentiment_summary = "情感分析失败"
@@ -542,7 +579,7 @@ class WeiboMultilingualSentimentAnalyzer:
                 "average_confidence": round(batch_result.average_confidence, 4),
                 "sentiment_distribution": sentiment_distribution,
                 "high_confidence_results": high_confidence_results,  # 返回所有高置信度结果，不做限制
-                "summary": sentiment_summary
+                "summary": sentiment_summary,
             }
         }
@@ -556,14 +593,32 @@ class WeiboMultilingualSentimentAnalyzer:
         return {
             "model_name": "tabularisai/multilingual-sentiment-analysis",
             "supported_languages": [
-                "中文", "英文", "西班牙文", "阿拉伯文", "日文", "韩文", 
-                "德文", "法文", "意大利文", "葡萄牙文", "俄文", "荷兰文",
-                "波兰文", "土耳其文", "丹麦文", "希腊文", "芬兰文", 
-                "瑞典文", "挪威文", "匈牙利文", "捷克文", "保加利亚文"
+                "中文",
+                "英文",
+                "西班牙文",
+                "阿拉伯文",
+                "日文",
+                "韩文",
+                "德文",
+                "法文",
+                "意大利文",
+                "葡萄牙文",
+                "俄文",
+                "荷兰文",
+                "波兰文",
+                "土耳其文",
+                "丹麦文",
+                "希腊文",
+                "芬兰文",
+                "瑞典文",
+                "挪威文",
+                "匈牙利文",
+                "捷克文",
+                "保加利亚文",
             ],
             "sentiment_levels": list(self.sentiment_map.values()),
             "is_initialized": self.is_initialized,
-            "device": str(self.device) if self.device else "未设置"
+            "device": str(self.device) if self.device else "未设置",
         }
@@ -576,13 +631,16 @@ def enable_sentiment_analysis() -> bool:
     return multilingual_sentiment_analyzer.enable()
-def disable_sentiment_analysis(reason: Optional[str] = None, drop_state: bool = False) -> None:
+def disable_sentiment_analysis(
+    reason: Optional[str] = None, drop_state: bool = False
+) -> None:
     """Public helper to disable sentiment analysis at runtime."""
     multilingual_sentiment_analyzer.disable(reason=reason, drop_state=drop_state)
-def analyze_sentiment(text_or_texts: Union[str, List[str]], 
-                     initialize_if_needed: bool = True) -> Union[SentimentResult, BatchSentimentResult]:
+def analyze_sentiment(
+    text_or_texts: Union[str, List[str]], initialize_if_needed: bool = True
+) -> Union[SentimentResult, BatchSentimentResult]:
     """
     便捷的情感分析函数
@@ -614,20 +672,26 @@ if __name__ == "__main__":
     if analyzer.initialize():
         # 测试单个文本
         result = analyzer.analyze_single_text("今天天气真好，心情特别棒！")
-        print(f"单个文本分析: {result.sentiment_label} (置信度: {result.confidence:.4f})")
+        print(
+            f"单个文本分析: {result.sentiment_label} (置信度: {result.confidence:.4f})"
+        )
         # 测试批量文本
         test_texts = [
             "这家餐厅的菜味道非常棒！",
             "服务态度太差了，很失望",
             "I absolutely love this product!",
-            "The customer service was disappointing."
+            "The customer service was disappointing.",
         ]
         batch_result = analyzer.analyze_batch(test_texts)
-        print(f"\n批量分析: 成功 {batch_result.success_count}/{batch_result.total_processed}")
+        print(
+            f"\n批量分析: 成功 {batch_result.success_count}/{batch_result.total_processed}"
+        )
         for result in batch_result.results:
-            print(f"'{result.text[:30]}...' -> {result.sentiment_label} ({result.confidence:.4f})")
+            print(
+                f"'{result.text[:30]}...' -> {result.sentiment_label} ({result.confidence:.4f})"
+            )
     else:
         print("模型初始化失败，无法进行测试")