243 lines
9.2 KiB
PHP
243 lines
9.2 KiB
PHP
<?php
|
||
|
||
namespace app\common;
|
||
|
||
use app\common\service\LLMService;
|
||
|
||
/**
|
||
* 参考文献类型识别器
|
||
*
|
||
* 识别顺序（分层，从权威到兜底）：
|
||
* 1. CrossRef 的 type 字段（若有 DOI 且命中）——最权威
|
||
* 2. 英文规则启发式（基于作者原文特征词/结构）
|
||
* 3. 大模型兜底（规则判为 other 或低置信度时）
|
||
* 4. 仍无法判定 → other
|
||
*
|
||
* 支持类型：journal / book / conference / thesis / web / other
|
||
*/
|
||
class ReferenceTypeClassifier
|
||
{
|
||
const TYPE_JOURNAL = 'journal';
|
||
const TYPE_BOOK = 'book';
|
||
const TYPE_CONFERENCE = 'conference';
|
||
const TYPE_THESIS = 'thesis';
|
||
const TYPE_WEB = 'web';
|
||
const TYPE_OTHER = 'other';
|
||
|
||
/** 是否允许使用大模型兜底 */
|
||
private $useLlm = true;
|
||
|
||
public function __construct($config = [])
|
||
{
|
||
if (is_array($config) && isset($config['use_llm'])) {
|
||
$this->useLlm = (bool)$config['use_llm'];
|
||
}
|
||
}
|
||
|
||
/**
|
||
* 主入口：根据作者原文（可选 CrossRef type）识别类型
|
||
*
|
||
* @param string $referContent 作者提供的原始参考文献字符串
|
||
* @param string $crossrefType CrossRef 返回的 type（可选）
|
||
* @return array ['type' => string, 'confidence' => float, 'source' => string]
|
||
*/
|
||
public function classify($referContent, $crossrefType = '')
|
||
{
|
||
$text = trim((string)$referContent);
|
||
|
||
// 1. CrossRef type 优先
|
||
$byCrossref = $this->mapCrossrefType($crossrefType);
|
||
if ($byCrossref !== '') {
|
||
return ['type' => $byCrossref, 'confidence' => 0.95, 'source' => 'crossref'];
|
||
}
|
||
|
||
if ($text === '') {
|
||
return ['type' => self::TYPE_OTHER, 'confidence' => 0.0, 'source' => 'fallback'];
|
||
}
|
||
|
||
// 2. 规则启发式
|
||
$byRule = $this->classifyByRule($text);
|
||
if ($byRule['type'] !== self::TYPE_OTHER && $byRule['confidence'] >= 0.8) {
|
||
return ['type' => $byRule['type'], 'confidence' => $byRule['confidence'], 'source' => 'rule'];
|
||
}
|
||
|
||
// 3. 大模型兜底
|
||
if ($this->useLlm) {
|
||
$byLlm = $this->classifyByLlm($text);
|
||
if ($byLlm !== '') {
|
||
return ['type' => $byLlm, 'confidence' => 0.7, 'source' => 'llm'];
|
||
}
|
||
}
|
||
|
||
// 规则给出的弱结果（若有）优先于纯兜底
|
||
if ($byRule['type'] !== self::TYPE_OTHER) {
|
||
return ['type' => $byRule['type'], 'confidence' => $byRule['confidence'], 'source' => 'rule'];
|
||
}
|
||
|
||
// 4. 兜底
|
||
return ['type' => self::TYPE_OTHER, 'confidence' => 0.0, 'source' => 'fallback'];
|
||
}
|
||
|
||
/**
|
||
* CrossRef type → 内部枚举映射；未命中返回空串
|
||
*/
|
||
/**
|
||
* 将细分类归并为排版用的三类：journal / book / other
|
||
*/
|
||
public function normalizeDispatchType($type)
|
||
{
|
||
$type = strtolower(trim((string)$type));
|
||
if ($type === self::TYPE_BOOK) {
|
||
return self::TYPE_BOOK;
|
||
}
|
||
if ($type === self::TYPE_JOURNAL) {
|
||
return self::TYPE_JOURNAL;
|
||
}
|
||
return self::TYPE_OTHER;
|
||
}
|
||
|
||
public function mapCrossrefType($crossrefType)
|
||
{
|
||
$t = strtolower(trim((string)$crossrefType));
|
||
if ($t === '') {
|
||
return '';
|
||
}
|
||
|
||
$map = [
|
||
// 期刊
|
||
'journal-article' => self::TYPE_JOURNAL,
|
||
'journal' => self::TYPE_JOURNAL,
|
||
'journal-volume' => self::TYPE_JOURNAL,
|
||
'journal-issue' => self::TYPE_JOURNAL,
|
||
// 图书
|
||
'book' => self::TYPE_BOOK,
|
||
'book-chapter' => self::TYPE_BOOK,
|
||
'book-part' => self::TYPE_BOOK,
|
||
'book-section' => self::TYPE_BOOK,
|
||
'book-set' => self::TYPE_BOOK,
|
||
'book-series' => self::TYPE_BOOK,
|
||
'book-track' => self::TYPE_BOOK,
|
||
'reference-book' => self::TYPE_BOOK,
|
||
'edited-book' => self::TYPE_BOOK,
|
||
'monograph' => self::TYPE_BOOK,
|
||
// 会议
|
||
'proceedings-article' => self::TYPE_CONFERENCE,
|
||
'proceedings' => self::TYPE_CONFERENCE,
|
||
'proceedings-series' => self::TYPE_CONFERENCE,
|
||
// 学位论文
|
||
'dissertation' => self::TYPE_THESIS,
|
||
// 在线/预印本
|
||
'posted-content' => self::TYPE_WEB,
|
||
];
|
||
|
||
return isset($map[$t]) ? $map[$t] : '';
|
||
}
|
||
|
||
/**
|
||
* 英文规则启发式：按「最独特 → 最普通」顺序判定
|
||
*
|
||
* @return array ['type' => string, 'confidence' => float]
|
||
*/
|
||
public function classifyByRule($text)
|
||
{
|
||
$hasDoi = (bool)preg_match('/\bdoi:\s*10\./i', $text) || (bool)preg_match('#doi\.org/#i', $text);
|
||
$hasUrl = (bool)preg_match('#https?://#i', $text);
|
||
// 期刊卷期页结构，如 2020;382(8):727-733 或 2020;10:100
|
||
// 年份与卷号之间容忍多余标点与全角符号（作者原文常见 "2026;, 44(5): 24-31"）
|
||
$hasJournalVol = (bool)preg_match(
|
||
'/\b(19|20)\d{2}[\s;:,，；：]+\d+\s*(\(\s*[^)]{1,12}\))?\s*[:：]\s*[A-Za-z]?\d+/u',
|
||
$text
|
||
);
|
||
|
||
// 1) 学位论文
|
||
if (preg_match('/\[(ph\.?d\.?|master(\'s)?|doctoral|masters)?\s*(thesis|dissertation)\]/i', $text)
|
||
|| preg_match('/\b(ph\.?d\.?|master\'?s|doctoral|doctorate)\b[^.]{0,40}\b(thesis|dissertation)\b/i', $text)
|
||
|| preg_match('/\b(thesis|dissertation)\b/i', $text)) {
|
||
return ['type' => self::TYPE_THESIS, 'confidence' => 0.9];
|
||
}
|
||
|
||
// 2) 会议论文
|
||
if (preg_match('/\bproceedings\b/i', $text)
|
||
|| preg_match('/\bin:\s*proc\b/i', $text)
|
||
|| preg_match('/\b(conference|symposium|workshop|congress)\b/i', $text)
|
||
|| preg_match('/\bannual meeting\b/i', $text)) {
|
||
return ['type' => self::TYPE_CONFERENCE, 'confidence' => 0.85];
|
||
}
|
||
|
||
// 3) 网页 / 在线资源
|
||
// "Available at:"/"Accessed" 同样出现在带 DOI 的期刊著录里（本系统自己的输出就带这个后缀），
|
||
// 故这些短语只在既无 DOI 也无卷期页结构时才作为网页依据
|
||
$hasWebPhrase = preg_match('/\[(internet|online)\]/i', $text)
|
||
|| preg_match('/\baccessed\b/i', $text)
|
||
|| preg_match('/\bavailable\s+(from|at)\b/i', $text)
|
||
|| preg_match('/\bcited\s+(19|20)\d{2}/i', $text);
|
||
if (($hasWebPhrase || $hasUrl) && !$hasDoi && !$hasJournalVol) {
|
||
return ['type' => self::TYPE_WEB, 'confidence' => 0.8];
|
||
}
|
||
|
||
// 4) 图书
|
||
if (preg_match('/\b\d+(st|nd|rd|th)\s+ed(ition)?\.?/i', $text)
|
||
|| preg_match('/\bisbn\b/i', $text)
|
||
|| preg_match('/\b(press|publisher|publishing house)\b/i', $text)
|
||
|| preg_match('/[A-Z][A-Za-z .]+:\s*[A-Z][A-Za-z .&]+;\s*(19|20)\d{2}/', $text)) {
|
||
// 期刊卷期结构更强时不判为书
|
||
if (!$hasJournalVol) {
|
||
return ['type' => self::TYPE_BOOK, 'confidence' => 0.8];
|
||
}
|
||
}
|
||
|
||
// 5) 期刊
|
||
if ($hasJournalVol || $hasDoi
|
||
|| preg_match('/\bvol\.?\s*\d+/i', $text)
|
||
|| preg_match('/\bpp?\.\s*\d+/i', $text)) {
|
||
return ['type' => self::TYPE_JOURNAL, 'confidence' => $hasJournalVol ? 0.85 : 0.6];
|
||
}
|
||
|
||
return ['type' => self::TYPE_OTHER, 'confidence' => 0.0];
|
||
}
|
||
|
||
/**
|
||
* 大模型兜底：只返回枚举值之一，失败返回空串
|
||
*/
|
||
public function classifyByLlm($text)
|
||
{
|
||
try {
|
||
$llm = new LLMService();
|
||
$system = 'You are a bibliography classifier. Classify the reference into exactly one type. '
|
||
. 'Allowed types: journal, book, conference, thesis, web, other. '
|
||
. 'journal=journal article; book=book or book chapter; conference=conference/proceedings paper; '
|
||
. 'thesis=dissertation/thesis; web=website or online resource; other=none of the above. '
|
||
. 'Respond with ONLY a JSON object: {"type":"<one of the allowed types>"}. No explanation.';
|
||
$user = "Reference:\n" . mb_substr($text, 0, 2000);
|
||
|
||
$content = $llm->requestChat([
|
||
['role' => 'system', 'content' => $system],
|
||
['role' => 'user', 'content' => $user],
|
||
], 0);
|
||
|
||
if ($content === null || $content === '') {
|
||
return '';
|
||
}
|
||
|
||
$parsed = $llm->parseJsonResponse($content);
|
||
$type = '';
|
||
if (is_array($parsed) && isset($parsed['type'])) {
|
||
$type = strtolower(trim((string)$parsed['type']));
|
||
} else {
|
||
// 兜底：直接从文本里找枚举词
|
||
if (preg_match('/\b(journal|book|conference|thesis|web|other)\b/i', $content, $m)) {
|
||
$type = strtolower($m[1]);
|
||
}
|
||
}
|
||
|
||
$allowed = [
|
||
self::TYPE_JOURNAL, self::TYPE_BOOK, self::TYPE_CONFERENCE,
|
||
self::TYPE_THESIS, self::TYPE_WEB, self::TYPE_OTHER,
|
||
];
|
||
return in_array($type, $allowed, true) ? $type : '';
|
||
} catch (\Throwable $e) {
|
||
return '';
|
||
}
|
||
}
|
||
}
|