Some checks failed
CI / PHP 8.3 / mariadb:10.6 (pull_request) Has been skipped
CI / PHP 8.4 / mariadb:10.6 (pull_request) Has been skipped
CI / PHP 8.3 / mariadb:11.4 (pull_request) Has been skipped
CI / PHP 8.4 / mariadb:11.4 (pull_request) Has been skipped
CI / PHP 8.3 / mysql:8.0 (pull_request) Has been skipped
CI / PHP 8.4 / mysql:8.0 (pull_request) Has been skipped
CI / PHP 8.3 / mysql:8.4 (pull_request) Has been skipped
CI / PHP 8.4 / mysql:8.4 (pull_request) Has been skipped
CI / Browser-Tests (pull_request) Failing after 19m38s
CI / PHP 8.3 / mysql:8.4-1 (pull_request) Successful in 24m17s
CI / Bekannte Sicherheitslücken in Paketen (pull_request) Successful in 9m13s
CI / Build-Artefakte aktuell (pull_request) Successful in 6m4s
Antworten mit Quellen aus Seiten, QM-Dokumenten, Glossar und Neuigkeiten – nur aus Sichtbarem, über den KI-Assistenten (Anthropic, OpenAI, Ollama). Wissensquellen als Registry für Plugins, Auswertung ohne Personenbezug. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
75 lines
3.4 KiB
PHP
75 lines
3.4 KiB
PHP
<?php
|
|
|
|
namespace App\Ask;
|
|
|
|
/**
|
|
* Suchbegriffe aus einer Frage in Alltagssprache: Füllwörter raus, Endungen grob
|
|
* abschneiden („beantrage“ findet „beantragt“ und „Beantragung“), höchstens acht.
|
|
*/
|
|
class Keywords
|
|
{
|
|
public const MAX = 8;
|
|
|
|
private const STOPWORDS = [
|
|
'aber', 'alle', 'allem', 'allen', 'aller', 'alles', 'also', 'als', 'am', 'an', 'auch', 'auf', 'aus', 'bei', 'beim', 'bin', 'bis',
|
|
'bitte', 'bist', 'da', 'dabei', 'dafür', 'dann', 'darf', 'darüber', 'das', 'dass', 'dazu', 'dein', 'deine', 'dem', 'den', 'denn',
|
|
'der', 'des', 'dich', 'die', 'dies', 'diese', 'diesem', 'diesen', 'dieser', 'doch', 'dort', 'du', 'durch', 'ein', 'eine', 'einem',
|
|
'einen', 'einer', 'eines', 'er', 'es', 'etwas', 'euch', 'für', 'gibt', 'gilt', 'habe', 'haben', 'hat', 'hier', 'ich', 'ihr', 'ihre',
|
|
'im', 'in', 'ist', 'ja', 'jetzt', 'kann', 'kannst', 'kein', 'keine', 'können', 'man', 'mein', 'meine', 'meinem', 'meinen', 'meiner',
|
|
'mich', 'mir', 'mit', 'muss', 'musst', 'müssen', 'nach', 'nicht', 'noch', 'nur', 'ob', 'oder', 'ohne', 'schon', 'sein', 'seine',
|
|
'sich', 'sie', 'sind', 'so', 'soll', 'sollte', 'über', 'um', 'und', 'uns', 'unser', 'unsere', 'unter', 'vom', 'von', 'vor', 'war',
|
|
'was', 'weiß', 'welche', 'welchem', 'welchen', 'welcher', 'welches', 'wem', 'wen', 'wenn', 'wer', 'werde', 'werden', 'wie', 'wieso',
|
|
'wieviel', 'wieviele', 'viel', 'viele', 'wir', 'wird', 'wo', 'woher', 'wohin', 'wann', 'warum', 'weshalb', 'zu', 'zum', 'zur',
|
|
'gerne', 'genau', 'eigentlich', 'immer', 'mal', 'mehr', 'sehr', 'einfach', 'bekomme', 'bekommen', 'machen', 'mache', 'finde',
|
|
'finden', 'gehen', 'geht', 'brauche', 'brauchen', 'stelle', 'wissen', 'frage', 'fragen', 'hallo', 'danke', 'portal',
|
|
];
|
|
|
|
/** Endungen, die abgeschnitten werden (längste zuerst), solange mindestens vier Zeichen bleiben */
|
|
private const ENDINGS = ['ungen', 'ung', 'en', 'er', 'es', 'e', 'n', 's'];
|
|
|
|
/**
|
|
* @return list<string>
|
|
*/
|
|
public static function from(string $question): array
|
|
{
|
|
$words = preg_split('/[^\p{L}\p{N}-]+/u', mb_strtolower($question), -1, PREG_SPLIT_NO_EMPTY) ?: [];
|
|
$terms = [];
|
|
|
|
foreach ($words as $word) {
|
|
$word = trim($word, '-');
|
|
|
|
if (mb_strlen($word) < 3 || in_array($word, self::STOPWORDS, true)) {
|
|
continue;
|
|
}
|
|
|
|
$terms[] = self::stem($word);
|
|
|
|
// „MAV-Wahl“ findet auch „Wahl der MAV“
|
|
if (str_contains($word, '-')) {
|
|
foreach (explode('-', $word) as $part) {
|
|
if (mb_strlen($part) >= 3 && ! in_array($part, self::STOPWORDS, true)) {
|
|
$terms[] = self::stem($part);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
return array_slice(array_values(array_unique($terms)), 0, self::MAX);
|
|
}
|
|
|
|
public static function stem(string $word): string
|
|
{
|
|
// Abkürzungen und Zahlen (z. B. „MAV“, „2026“) unverändert lassen
|
|
if (mb_strlen($word) <= 4 || preg_match('/\d/u', $word)) {
|
|
return $word;
|
|
}
|
|
|
|
foreach (self::ENDINGS as $ending) {
|
|
if (str_ends_with($word, $ending) && mb_strlen($word) - mb_strlen($ending) >= 4) {
|
|
return mb_substr($word, 0, mb_strlen($word) - mb_strlen($ending));
|
|
}
|
|
}
|
|
|
|
return $word;
|
|
}
|
|
}
|