多语言查询优化

tangwang
1 parent 11237cf2
Showing 6 changed files with 126 additions and 29 deletions Show diff stats
frontend/index.html
frontend/static/js/app.js
query/query_parser.py
query/translator.py
search/es_query_builder.py
search/searcher.py
@@ -142,6 +142,6 @@
         <p>SearchEngine © 2025 | API: <span id="apiUrl">Loading...</span></p>
     </footer>
-    <script src="/static/js/app.js?v=3.7"></script>
+    <script src="/static/js/app.js?v=3.1"></script>
 </body>
 </html>
@@ -554,7 +554,7 @@ function displayDebugInfo(data) {
         if (data.query_info) {
             let html = '<div style="padding: 10px;">';
             html += `<div><strong>original_query:</strong> ${escapeHtml(data.query_info.original_query || 'N/A')}</div>`;
-            html += `<div><strong>detected_language:</strong> ${getLanguageName(data.query_info.detected_language)}</div>`;
+            html += `<div><strong>detected_language:</strong> ${data.query_info.detected_languag}</div>`;
             html += '</div>';
             debugInfoDiv.innerHTML = html;
         } else {
@@ -573,7 +573,7 @@ function displayDebugInfo(data) {
         html += `<div>original_query: ${escapeHtml(debugInfo.query_analysis.original_query || 'N/A')}</div>`;
         html += `<div>normalized_query: ${escapeHtml(debugInfo.query_analysis.normalized_query || 'N/A')}</div>`;
         html += `<div>rewritten_query: ${escapeHtml(debugInfo.query_analysis.rewritten_query || 'N/A')}</div>`;
-        html += `<div>detected_language: ${getLanguageName(debugInfo.query_analysis.detected_language)}</div>`;
+        html += `<div>detected_language: ${debugInfo.query_analysis.detected_language}</div>`;
         html += `<div>domain: ${escapeHtml(debugInfo.query_analysis.domain || 'default')}</div>`;
         html += `<div>is_simple_query: ${debugInfo.query_analysis.is_simple_query ? 'yes' : 'no'}</div>`;
@@ -581,7 +581,7 @@ function displayDebugInfo(data) {
             html += '<div>translations: ';
             for (const [lang, translation] of Object.entries(debugInfo.query_analysis.translations)) {
                 if (translation) {
-                    html += `${getLanguageName(lang)}: ${escapeHtml(translation)}; `;
+                    html += `${lang}: ${escapeHtml(translation)}; `;
                 }
             }
             html += '</div>';
@@ -669,14 +669,3 @@ function formatDate(dateStr) {
     }
 }
-function getLanguageName(code) {
-    const names = {
-        'zh': '中文',
-        'en': 'English',
-        'ru': 'Русский',
-        'ar': 'العربية',
-        'ja': '日本語',
-        'unknown': 'Unknown'
-    };
-    return names[code] || code;
-}
@@ -27,7 +27,7 @@ class ParsedQuery:
         original_query: str,
         normalized_query: str,
         rewritten_query: Optional[str] = None,
-        detected_language: str = "unknown",
+        detected_language: Optional[str] = None,
         translations: Dict[str, str] = None,
         query_vector: Optional[np.ndarray] = None,
         domain: str = "default",
@@ -210,6 +210,9 @@ class QueryParser:
         # Stage 3: Language detection
         detected_lang = self.language_detector.detect(query_text)
+        # Use default language if detection failed (None or "unknown")
+        if not detected_lang or detected_lang == "unknown":
+            detected_lang = self.config.query_config.default_language
         log_info(f"语言检测 | 检测到语言: {detected_lang}")
         if context:
             context.store_intermediate_result('detected_language', detected_lang)
@@ -12,6 +12,7 @@ https://developers.deepl.com/api-reference/translate/request-translation
 """
 import requests
+import re
 from concurrent.futures import ThreadPoolExecutor
 from typing import Dict, List, Optional
 from utils.cache import DictCache
@@ -110,6 +111,15 @@ class Translator:
         if source_lang:
             source_lang = source_lang.lower()
+        # Optimization: Skip translation if not needed
+        if target_lang == 'en' and self._is_english_text(text):
+            logger.debug(f"[Translator] Text is already English, skipping translation: '{text[:50]}...'")
+            return text
+        
+        if target_lang == 'zh' and (self._contains_chinese(text) or self._is_pure_number(text)):
+            logger.debug(f"[Translator] Text contains Chinese or is pure number, skipping translation: '{text[:50]}...'")
+            return text
+
         # Use provided context or default context
         translation_context = context or self.translation_context
@@ -312,8 +322,10 @@ class Translator:
         In async_mode=True (default):
         - Returns cached translations immediately if available
-        - Launches async tasks for missing translations (non-blocking)
-        - Returns None for missing translations (will be available in cache next time)
+        - For translations that can be optimized (e.g., pure numbers, already in target language),
+          returns result immediately via synchronous call
+        - Launches async tasks for other missing translations (non-blocking)
+        - Returns None for missing translations that require async processing
         In async_mode=False:
         - Waits for all translations to complete (blocking)
@@ -331,6 +343,7 @@ class Translator:
         """
         results = {}
         missing_langs = []
+        async_langs = []
         # First, get cached translations
         for lang in target_langs:
@@ -340,13 +353,32 @@ class Translator:
             else:
                 missing_langs.append(lang)
-        # If async mode and there are missing translations, launch async tasks
+        # If async mode and there are missing translations
         if async_mode and missing_langs:
+            # Check if translation can be optimized (immediate return)
             for lang in missing_langs:
-                self._translate_async(text, lang, source_lang, context, prompt)
-            # Return None for missing translations
-            for lang in missing_langs:
-                results[lang] = None
+                target_lang = lang.lower()
+                # Check optimization conditions (same as in translate method)
+                can_optimize = False
+                if target_lang == 'en' and self._is_english_text(text):
+                    can_optimize = True
+                elif target_lang == 'zh' and (self._contains_chinese(text) or self._is_pure_number(text)):
+                    can_optimize = True
+                
+                if can_optimize:
+                    # Can be optimized, call translate synchronously for immediate result
+                    results[lang] = self.translate(text, lang, source_lang, context, prompt)
+                else:
+                    # Requires actual translation, add to async list
+                    async_langs.append(lang)
+            
+            # Launch async tasks for translations that require actual API calls
+            if async_langs:
+                for lang in async_langs:
+                    self._translate_async(text, lang, source_lang, context, prompt)
+                # Return None for async translations
+                for lang in async_langs:
+                    results[lang] = None
         else:
             # Synchronous mode: wait for all translations
             for lang in missing_langs:
@@ -496,3 +528,67 @@ class Translator:
         # Otherwise, translate to all supported languages
         return supported_langs
+    
+    def _is_english_text(self, text: str) -> bool:
+        """
+        Check if text is primarily English (ASCII letters, numbers, common punctuation).
+        
+        Args:
+            text: Text to check
+            
+        Returns:
+            True if text appears to be English
+        """
+        if not text or not text.strip():
+            return True
+        
+        # Remove whitespace and common punctuation
+        text_clean = re.sub(r'[\s\.,!?;:\-\'\"\(\)\[\]{}]', '', text)
+        if not text_clean:
+            return True
+        
+        # Check if all remaining characters are ASCII (letters, numbers)
+        # This is a simple heuristic: if most characters are ASCII, it's likely English
+        ascii_count = sum(1 for c in text_clean if ord(c) < 128)
+        ratio = ascii_count / len(text_clean) if text_clean else 0
+        
+        # If more than 80% are ASCII characters, consider it English
+        return ratio > 0.8
+    
+    def _contains_chinese(self, text: str) -> bool:
+        """
+        Check if text contains Chinese characters (Han characters).
+        
+        Args:
+            text: Text to check
+            
+        Returns:
+            True if text contains Chinese characters
+        """
+        if not text:
+            return False
+        
+        # Check for Chinese characters (Unicode range: \u4e00-\u9fff)
+        chinese_pattern = re.compile(r'[\u4e00-\u9fff]')
+        return bool(chinese_pattern.search(text))
+    
+    def _is_pure_number(self, text: str) -> bool:
+        """
+        Check if text is purely numeric (digits, possibly with spaces, dots, commas).
+        
+        Args:
+            text: Text to check
+            
+        Returns:
+            True if text is purely numeric
+        """
+        if not text or not text.strip():
+            return False
+        
+        # Remove whitespace, dots, commas (common number separators)
+        text_clean = re.sub(r'[\s\.,]', '', text.strip())
+        if not text_clean:
+            return False
+        
+        # Check if all remaining characters are digits
+        return text_clean.isdigit()
@@ -25,7 +25,8 @@ class ESQueryBuilder:
         image_embedding_field: Optional[str] = None,
         source_fields: Optional[List[str]] = None,
         function_score_config: Optional[FunctionScoreConfig] = None,
-        enable_multilang_search: bool = True
+        enable_multilang_search: bool = True,
+        default_language: str = "zh"
     ):
         """
         Initialize query builder.
@@ -38,6 +39,7 @@ class ESQueryBuilder:
             source_fields: Fields to return in search results (_source includes)
             function_score_config: Function score configuration
             enable_multilang_search: Enable multi-language search using translations
+            default_language: Default language to use when detection fails or returns "unknown"
         """
         self.index_name = index_name
         self.match_fields = match_fields
@@ -46,6 +48,7 @@ class ESQueryBuilder:
         self.source_fields = source_fields
         self.function_score_config = function_score_config
         self.enable_multilang_search = enable_multilang_search
+        self.default_language = default_language
     def _split_filters_for_faceting(
         self,
@@ -422,7 +425,7 @@ class ESQueryBuilder:
         # Get query analysis from parsed_query
         translations = {}
-        language = 'zh'
+        language = self.default_language
         keywords = ""
         token_count = 0
         is_short_query = False
@@ -430,7 +433,12 @@ class ESQueryBuilder:
         if parsed_query:
             translations = parsed_query.translations or {}
-            language = parsed_query.detected_language or 'zh'
+            # Use default language if detected_language is None or "unknown"
+            detected_lang = parsed_query.detected_language
+            if not detected_lang or detected_lang == "unknown":
+                language = self.default_language
+            else:
+                language = detected_lang
             keywords = getattr(parsed_query, 'keywords', '') or ""
             token_count = getattr(parsed_query, 'token_count', 0) or 0
             is_short_query = getattr(parsed_query, 'is_short_query', False)
@@ -458,7 +466,7 @@ class ESQueryBuilder:
         # 2. Translation queries - lower boost (0.4) for other languages
         if self.enable_multilang_search:
-            if language != 'zh' and translations.get('zh') and translations['zh'] != query_text:
+            if language != 'zh' and translations.get('zh'):
                 zh_fields, _ = self._get_match_fields('zh')
                 should_clauses.append({
                     "multi_match": {
@@ -472,7 +480,7 @@ class ESQueryBuilder:
                     }
                 })
-            if language != 'en' and translations.get('en') and translations['en'] != query_text:
+            if language != 'en' and translations.get('en'):
                 en_fields, _ = self._get_match_fields('en')
                 should_clauses.append({
                     "multi_match": {
@@ -113,7 +113,8 @@ class Searcher:
             image_embedding_field=self.image_embedding_field,
             source_fields=self.source_fields,
             function_score_config=self.config.function_score,
-            enable_multilang_search=self.config.query_config.enable_multilang_search
+            enable_multilang_search=self.config.query_config.enable_multilang_search,
+            default_language=self.config.query_config.default_language
         )
     def search(