参考文献作者堆叠

参考文献相关性检测
作者ai写作辅助检测工作

application/common/ReferenceCheckService.php

@@ -2345,7 +2345,7 @@ class ReferenceCheckService
         }
 
         $slice = $this->buildCitationContextText($raw, $extendedStart, $textEnd);
-        $slice = ltrim($slice, ". \t\n\r");
+        $slice = ltrim($slice, "., \t\n\r");
         if (trim($slice) === '') {
             return $fallback;
         }
@@ -3894,9 +3894,13 @@ class ReferenceCheckService
         $hasPriorCiteInParagraph = ($prevTagEnd > $paragraphStart);
         $sentenceStart = $this->findSentenceStart($content, $tagStart);
 
-        // 段内首个引用：整段到标签前；后续引用：取「本句」起点（可早于上一标签），避免只剩 “and external environment” 再误用标签后文本
+        // 段内首个引用：整段到标签前；后续引用：不早于上一标签结束，并可向前扩展若干句覆盖紧邻 claim
         if ($hasPriorCiteInParagraph) {
-            $localStart = max($paragraphStart, $sentenceStart);
+            $anchor = max($prevTagEnd, $sentenceStart);
+            $localStart = $this->extendContextStartBackward($content, $anchor, $prevTagEnd, 2);
+            if ($localStart >= $prevTagEnd) {
+                $localStart = $this->advancePastPriorCitationBoundary($content, $localStart);
+            }
         } else {
             $localStart = $this->capContextStartBeforeTag($content, $tagStart, $paragraphStart);
         }
@@ -3904,19 +3908,34 @@ class ReferenceCheckService
         // 默认：引用标签前的论述
         $localEnd = $tagStart;
         $originalText = $this->buildCitationContextText($content, $localStart, $localEnd);
+        $before = $originalText;
+        $isAuthorOnly = $this->isAuthorOnlyLeadIn($before);
 
-        // 仅段内首个引用、且标签前极短（如句末 ICU nurses [14]）时，才改用标签后片段；同段多引禁止标签后截取（会错取下一句）
-        $allowTrailing = !$hasPriorCiteInParagraph;
-        if ($allowTrailing && (
-            !$this->isMeaningfulCitationContext($originalText)
-            || $this->shouldUseTrailingCitationContext($content, $localStart, $tagStart, $tagEnd)
-        )) {
+        // 作者缩写引用（Chen [33]、Zou [24] found that...）：向前扩展到段落/句群，必要时并入标签后叙述
+        if ($isAuthorOnly) {
+            $localStart = $this->capContextStartBeforeTag($content, $tagStart, $paragraphStart);
             $trailEnd = ($nextTagStart < $sentenceEnd) ? $nextTagStart : $sentenceEnd;
             $trailText = $this->buildCitationContextText($content, $tagEnd, $trailEnd);
             if ($this->isMeaningfulCitationContext($trailText)) {
-                $localStart = $tagEnd;
                 $localEnd = $trailEnd;
-                $originalText = $trailText;
+            } else {
+                $localEnd = $tagStart;
+            }
+            $originalText = $this->buildCitationContextText($content, $localStart, $localEnd);
+        } else {
+            // 仅段内首个引用、且标签前极短时，才改用标签后片段；同段多引禁止（会错取下一句）
+            $allowTrailing = !$hasPriorCiteInParagraph;
+            if ($allowTrailing && (
+                !$this->isMeaningfulCitationContext($originalText)
+                || $this->shouldUseTrailingCitationContext($content, $localStart, $tagStart, $tagEnd)
+            )) {
+                $trailEnd = ($nextTagStart < $sentenceEnd) ? $nextTagStart : $sentenceEnd;
+                $trailText = $this->buildCitationContextText($content, $tagEnd, $trailEnd);
+                if ($this->isMeaningfulCitationContext($trailText)) {
+                    $localStart = $tagEnd;
+                    $localEnd = $trailEnd;
+                    $originalText = $trailText;
+                }
             }
         }
 
@@ -3934,6 +3953,22 @@ class ReferenceCheckService
         return [$localStart, $localEnd, $originalText];
     }
 
+    /**
+     * 标签前仅有作者姓氏/缩写（如 Chen、Zou）时视为作者引导引用，需扩展上下文。
+     */
+    private function isAuthorOnlyLeadIn($text)
+    {
+        $text = trim((string)$text);
+        if ($text === '') {
+            return false;
+        }
+        if (mb_strlen($text) < 25) {
+            return true;
+        }
+
+        return preg_match('/^[A-Z][a-zA-Z\'\-]{0,24}\.?$/u', $text) === 1;
+    }
+
     /**
      * 标签前仅有作者缩写等极短片段时，改用标签后上下文
      */
@@ -4097,6 +4132,27 @@ class ReferenceCheckService
         return $start;
     }
 
+    /**
+     * 上一引用标签结束后，跳过空白及紧随其后的句末/分句标点（如 "[1]. And"、"[2], accounting"）
+     */
+    private function advancePastPriorCitationBoundary($content, $pos)
+    {
+        $pos = intval($pos);
+        $len = strlen($content);
+        while ($pos < $len) {
+            while ($pos < $len && ($content[$pos] === ' ' || $content[$pos] === "\t" || $content[$pos] === "\n" || $content[$pos] === "\r")) {
+                $pos++;
+            }
+            if ($pos < $len && ($content[$pos] === '.' || $content[$pos] === ',')) {
+                $pos++;
+                continue;
+            }
+            break;
+        }
+
+        return $pos;
+    }
+
     /**
      * 过滤仅标点、过短或无字母/汉字的上下文（如去掉标签后只剩 "."）
      */