fix: improve Chinese NLP filetype detection accuracy

autotests/dfm-search-tests/tst_chinese_nlp.cpp

@@ -168,6 +168,7 @@ private Q_SLOTS:
     void fileType_archive_allSynonyms();
     void fileType_application_allSynonyms();
     void fileType_design_source_allSynonyms();
+    void fileType_designSource_shadowsImageSubstring();
 
     // Filetype specific+general combination tests (span leak regression)
     void fileType_specificAndGeneral_pptDocument();
@@ -1287,10 +1288,9 @@ void tst_ChineseNLP::fileType_archive_allSynonyms()
 void tst_ChineseNLP::fileType_application_allSynonyms()
 {
     // Requirements: 安装包, 软件, 应用, 脚本, 程序, 包
-    // NOTE: "包" excluded from rules to avoid false positives with "表情包", "压缩包"
     const QStringList inputs = {
         QStringLiteral("安装包"), QStringLiteral("软件"), QStringLiteral("应用"),
-        QStringLiteral("脚本"), QStringLiteral("程序")
+        QStringLiteral("脚本"), QStringLiteral("程序"), QStringLiteral("包")
     };
     for (const QString &input : inputs) {
         ParsedIntent intent;
@@ -1320,6 +1320,28 @@ void tst_ChineseNLP::fileType_design_source_allSynonyms()
     }
 }
 
+void tst_ChineseNLP::fileType_designSource_shadowsImageSubstring()
+{
+    ParsedIntent intent;
+    m_parser->parse(QStringLiteral("昨天创建的矢量图"), intent);
+
+    QVERIFY2(setEquals(intent.fileExtensions(), expectedExtsForRule("filetype_design_source")),
+             qPrintable(QStringLiteral("Extensions: ") + intent.fileExtensions().join(',')));
+
+    bool hasDesignSource = false;
+    bool hasImage = false;
+    for (const MatchSpan &span : intent.consumedSpans()) {
+        if (span.ruleId() == QLatin1String("filetype_design_source")) {
+            hasDesignSource = true;
+        } else if (span.ruleId() == QLatin1String("filetype_image")) {
+            hasImage = true;
+        }
+    }
+
+    QVERIFY(hasDesignSource);
+    QVERIFY(!hasImage);
+}
+
 // ===== Filetype specific+general combination tests (span leak regression) =====
 // 当 specific filetype 规则与 general filetype 规则同时命中时，general 规则匹配的
 // 文本（文档/表格/幻灯片）必须被消费，不得泄漏成 keyword；且 fileExtensions 应只

src/dfm-search/dfm-search-lib/semantic/extractors/filetypeextractor.cpp

@@ -48,7 +48,6 @@ void FileTypeExtractor::extract(const QString &input, ParsedIntent &intent)
             continue;
         }
 
-        // Always consume the matched span so the matched text doesn't leak into keywords
         MatchSpan span;
         span.setStart(m.capturedStart());
         span.setEnd(m.capturedEnd());

src/dfm-search/dfm-search-lib/semantic/rules/zh_CN/filetype_rules.json

@@ -129,7 +129,7 @@
                 },
                 {
                     "id": "filetype_application",
-                    "pattern": "安装包|软件|应用|脚本|程序|包",
+                    "pattern": "安装包|软件|应用|脚本|程序|包(?!含)",
                     "description": "Application packages",
                     "enabled": true,
                     "priority": 150,
@@ -143,7 +143,7 @@
                     "pattern": "源文件|设计稿|矢量图|工程文件|psd|fig|sketch",
                     "description": "Design source files",
                     "enabled": true,
-                    "priority": 150,
+                    "priority": 160,
                     "metadata": {
                         "extensions": ["psd", "ai", "fig", "sketch"]
                     }