|
@@ -923,19 +923,25 @@ class OpenSearchService
|
|
|
/**
|
|
/**
|
|
|
* definition 的计分表(黑体字部分,每个原始词各算一次)
|
|
* definition 的计分表(黑体字部分,每个原始词各算一次)
|
|
|
*
|
|
*
|
|
|
- * - single 该词是单独黑体字,拼写相同
|
|
|
|
|
- * - single_synonym 该词经同义词可替代地是单独黑体字(拼写相同的也会命中)
|
|
|
|
|
- * - multi 该词出现在多词黑体短语里,拼写相同
|
|
|
|
|
- * - multi_phrase 整句按原顺序出现在多词黑体短语里(≥2 个词才有)
|
|
|
|
|
|
|
+ * 主分不分词形:同义词词典的一行就是一个词的全部变格变位,所以「任一词形」走同义词,
|
|
|
|
|
+ * 与 v2 search-pali-wbw 把全部词形一视同仁的口径一致;原拼写只给少量加分,排在同档前面。
|
|
|
*
|
|
*
|
|
|
|
|
+ * - single 该词(任一词形)是单独黑体字
|
|
|
|
|
+ * - single_exact 再加:拼写也相同
|
|
|
|
|
+ * - multi 该词(任一词形)出现在多词黑体短语里
|
|
|
|
|
+ * - multi_exact 再加:拼写也相同
|
|
|
|
|
+ * - multi_phrase 整句按原顺序、拼写相同出现在多词黑体短语里(≥2 个词才有)
|
|
|
|
|
+ *
|
|
|
|
|
+ * 每个词:单独黑体 ≥ 100 > 多词黑体 ≤ 35,于是命中的单独黑体字越多排越前。
|
|
|
* 正文得分按 DEFINITION_TEXT_SCALE 缩小后叠加,只用来给黑体同分的文档排先后。
|
|
* 正文得分按 DEFINITION_TEXT_SCALE 缩小后叠加,只用来给黑体同分的文档排先后。
|
|
|
*
|
|
*
|
|
|
- * @var array{single: int, single_synonym: int, multi: int, multi_phrase: int}
|
|
|
|
|
|
|
+ * @var array{single: int, single_exact: int, multi: int, multi_exact: int, multi_phrase: int}
|
|
|
*/
|
|
*/
|
|
|
private const DEFINITION_SCORES = [
|
|
private const DEFINITION_SCORES = [
|
|
|
'single' => 100,
|
|
'single' => 100,
|
|
|
- 'single_synonym' => 20,
|
|
|
|
|
|
|
+ 'single_exact' => 10,
|
|
|
'multi' => 30,
|
|
'multi' => 30,
|
|
|
|
|
+ 'multi_exact' => 5,
|
|
|
'multi_phrase' => 30,
|
|
'multi_phrase' => 30,
|
|
|
];
|
|
];
|
|
|
|
|
|
|
@@ -1134,8 +1140,11 @@ class OpenSearchService
|
|
|
* 义注里的黑体字通常就是被解释的词,所以黑体字命中决定排序,正文只做次级排序:
|
|
* 义注里的黑体字通常就是被解释的词,所以黑体字命中决定排序,正文只做次级排序:
|
|
|
* - 每个原始词单独算:命中的单独黑体字(bold_single)越多分越高
|
|
* - 每个原始词单独算:命中的单独黑体字(bold_single)越多分越高
|
|
|
* - 单独黑体字 > 多词黑体短语(bold_multi)里的词
|
|
* - 单独黑体字 > 多词黑体短语(bold_multi)里的词
|
|
|
- * - 拼写相同 > 同义词可替代:bold_* 建索引用 standard,查询时显式指定
|
|
|
|
|
- * analyzer=standard 即是「拼写相同」,不指定则走带同义词的 pali_query_analyzer
|
|
|
|
|
|
|
+ * - 词形不分高下(走同义词,即该词的全部变格变位),原拼写只在同档内加少量分:
|
|
|
|
|
+ * bold_* 建索引用 standard,查询时显式指定 analyzer=standard 即是「拼写相同」,
|
|
|
|
|
+ * 不指定则走带同义词的 pali_query_analyzer
|
|
|
|
|
+ * - 多词黑体也按任一词形算:bold_* 取自 wbw_templates.real(已拆连音),
|
|
|
|
|
+ * 正文里粘了连音(如 vitakkanti)匹配不到的词,往往靠它召回
|
|
|
* - 用 match_phrase 而不是 match:中文在 standard 下按字切分,phrase 要求字连续,
|
|
* - 用 match_phrase 而不是 match:中文在 standard 下按字切分,phrase 要求字连续,
|
|
|
* 避免一个字重合就命中
|
|
* 避免一个字重合就命中
|
|
|
* 计分表见 DEFINITION_SCORES;正文部分复用 fuzzy 的计分,乘 DEFINITION_TEXT_SCALE。
|
|
* 计分表见 DEFINITION_SCORES;正文部分复用 fuzzy 的计分,乘 DEFINITION_TEXT_SCALE。
|
|
@@ -1155,18 +1164,16 @@ class OpenSearchService
|
|
|
|
|
|
|
|
// 停用词不单独算黑体分(整句的 multi_phrase 仍带上它们)
|
|
// 停用词不单独算黑体分(整句的 multi_phrase 仍带上它们)
|
|
|
foreach ($this->scoringWords($words) as $word) {
|
|
foreach ($this->scoringWords($words) as $word) {
|
|
|
- $should[] = $this->constantScore(
|
|
|
|
|
- ['match_phrase' => ['bold_single' => ['query' => $word, 'analyzer' => 'standard']]],
|
|
|
|
|
- self::DEFINITION_SCORES['single'],
|
|
|
|
|
- );
|
|
|
|
|
- $should[] = $this->constantScore(
|
|
|
|
|
- ['match_phrase' => ['bold_single' => $word]],
|
|
|
|
|
- self::DEFINITION_SCORES['single_synonym'],
|
|
|
|
|
- );
|
|
|
|
|
- $should[] = $this->constantScore(
|
|
|
|
|
- ['match_phrase' => ['bold_multi' => ['query' => $word, 'analyzer' => 'standard']]],
|
|
|
|
|
- self::DEFINITION_SCORES['multi'],
|
|
|
|
|
- );
|
|
|
|
|
|
|
+ foreach (['single' => 'bold_single', 'multi' => 'bold_multi'] as $tier => $field) {
|
|
|
|
|
+ $should[] = $this->constantScore(
|
|
|
|
|
+ ['match_phrase' => [$field => $word]],
|
|
|
|
|
+ self::DEFINITION_SCORES[$tier],
|
|
|
|
|
+ );
|
|
|
|
|
+ $should[] = $this->constantScore(
|
|
|
|
|
+ ['match_phrase' => [$field => ['query' => $word, 'analyzer' => 'standard']]],
|
|
|
|
|
+ self::DEFINITION_SCORES[$tier.'_exact'],
|
|
|
|
|
+ );
|
|
|
|
|
+ }
|
|
|
}
|
|
}
|
|
|
|
|
|
|
|
if (count($words) >= 2) {
|
|
if (count($words) >= 2) {
|