| 1 |
<?php |
| 2 |
|
| 3 |
if (!defined('ABSPATH')) { |
| 4 |
exit; |
| 5 |
} |
| 6 |
|
| 7 |
/** |
| 8 |
* N-gram set similarity primitives. |
| 9 |
* |
| 10 |
* Owns the Dice coefficient (set-overlap similarity) and the proximity-merge |
| 11 |
* routine that interleaves two pre-sorted result arrays by distance to a |
| 12 |
* target ngram count. Pure computation; no I/O. |
| 13 |
* |
| 14 |
* Used by NGramFilter::findSimilarPages() during candidate selection and by |
| 15 |
* NGramCacheRepository::getCachedNGramsFiltered() to merge below/above-target |
| 16 |
* query halves. Algorithm tests exercise these directly. |
| 17 |
*/ |
| 18 |
class ABJ_404_Solution_NGramSimilarity { |
| 19 |
|
| 20 |
/** |
| 21 |
* Compute Dice coefficient similarity between two N-gram sets. |
| 22 |
* |
| 23 |
* Dice coefficient: 2 * |intersection| / (|set1| + |set2|) |
| 24 |
* Range: 0.0 (no overlap) to 1.0 (identical). |
| 25 |
* |
| 26 |
* Threshold correlation: |
| 27 |
* - 0.4 = ~30% edit distance (recommended) |
| 28 |
* - 0.5 = ~20% edit distance |
| 29 |
* - 0.6 = ~10% edit distance |
| 30 |
* |
| 31 |
* @param array{bi?: array<int, string>, tri?: array<int, string>} $ngrams1 |
| 32 |
* @param array{bi?: array<int, string>, tri?: array<int, string>} $ngrams2 |
| 33 |
* @return float Similarity score 0.0 to 1.0 |
| 34 |
*/ |
| 35 |
public function diceCoefficient($ngrams1, $ngrams2) { |
| 36 |
$set1 = array_merge( |
| 37 |
isset($ngrams1['bi']) ? $ngrams1['bi'] : [], |
| 38 |
isset($ngrams1['tri']) ? $ngrams1['tri'] : [] |
| 39 |
); |
| 40 |
$set2 = array_merge( |
| 41 |
isset($ngrams2['bi']) ? $ngrams2['bi'] : [], |
| 42 |
isset($ngrams2['tri']) ? $ngrams2['tri'] : [] |
| 43 |
); |
| 44 |
|
| 45 |
if (empty($set1) || empty($set2)) { |
| 46 |
return 0.0; |
| 47 |
} |
| 48 |
|
| 49 |
$set1 = array_flip($set1); |
| 50 |
$set2 = array_flip($set2); |
| 51 |
|
| 52 |
$intersection = count(array_intersect_key($set1, $set2)); |
| 53 |
|
| 54 |
// Defensive: empty() check above should guarantee denominator > 0. |
| 55 |
$denominator = count($set1) + count($set2); |
| 56 |
return ($denominator > 0) ? (2.0 * $intersection) / $denominator : 0.0; |
| 57 |
} |
| 58 |
|
| 59 |
/** |
| 60 |
* Merge two arrays sorted by proximity to target, interleaving results. |
| 61 |
* |
| 62 |
* Both input arrays must be pre-sorted by proximity to the target: |
| 63 |
* - $below: ngram_count <= target, ordered DESC by ngram_count (closest first) |
| 64 |
* - $above: ngram_count > target, ordered ASC by ngram_count (closest first) |
| 65 |
* |
| 66 |
* @param array<int, mixed> $below |
| 67 |
* @param array<int, mixed> $above |
| 68 |
* @param int $targetNgramCount |
| 69 |
* @param int $limit |
| 70 |
* @return array<int, mixed> |
| 71 |
*/ |
| 72 |
public function mergeByProximity($below, $above, $targetNgramCount, $limit) { |
| 73 |
$result = []; |
| 74 |
$i = 0; |
| 75 |
$j = 0; |
| 76 |
$belowCount = count($below); |
| 77 |
$aboveCount = count($above); |
| 78 |
|
| 79 |
while (count($result) < $limit && ($i < $belowCount || $j < $aboveCount)) { |
| 80 |
$belowEntry = $below[$i] ?? null; |
| 81 |
$aboveEntry = $above[$j] ?? null; |
| 82 |
$belowNgramRaw = (is_array($belowEntry) && isset($belowEntry['ngram_count'])) ? $belowEntry['ngram_count'] : 0; |
| 83 |
$belowNgramCount = is_scalar($belowNgramRaw) ? (int)$belowNgramRaw : 0; |
| 84 |
$aboveNgramRaw = (is_array($aboveEntry) && isset($aboveEntry['ngram_count'])) ? $aboveEntry['ngram_count'] : 0; |
| 85 |
$aboveNgramCount = is_scalar($aboveNgramRaw) ? (int)$aboveNgramRaw : 0; |
| 86 |
$distBelow = ($i < $belowCount) |
| 87 |
? abs($belowNgramCount - $targetNgramCount) |
| 88 |
: PHP_INT_MAX; |
| 89 |
$distAbove = ($j < $aboveCount) |
| 90 |
? abs($aboveNgramCount - $targetNgramCount) |
| 91 |
: PHP_INT_MAX; |
| 92 |
|
| 93 |
// Prefer below on tie (includes exact matches) |
| 94 |
if ($distBelow <= $distAbove) { |
| 95 |
$result[] = $below[$i]; |
| 96 |
$i++; |
| 97 |
} else { |
| 98 |
$result[] = $above[$j]; |
| 99 |
$j++; |
| 100 |
} |
| 101 |
} |
| 102 |
|
| 103 |
return $result; |
| 104 |
} |
| 105 |
} |
| 106 |
|
