/**
* BM25 알고리즘 구현
* 한국어 문서 검색을 위한 확률적 랭킹 함수
*/
export interface DocumentChunk {
id: number;
chunkId: number;
originTitle: string;
text: string;
wordCount: number;
}
export interface SearchResult {
id: number;
chunkId: number;
score: number;
}
export class BM25Calculator {
private readonly allChunks: DocumentChunk[];
private readonly totalCount: number;
private readonly averageDocLength: number;
private readonly N: number;
constructor(
chunks: DocumentChunk[],
private readonly k1: number = 1.2,
private readonly b: number = 0.75
) {
this.allChunks = chunks;
this.totalCount = chunks.reduce((count, chunk) => count + chunk.wordCount, 0);
this.averageDocLength = this.totalCount / chunks.length;
this.N = chunks.length;
}
/**
* 주어진 키워드로 문서 검색
* @param keywords 검색할 키워드 (정규식 패턴)
* @returns 점수가 높은 순으로 정렬된 검색 결과
*/
calculate(keywords: string): SearchResult[] {
const { termFrequencies, docFrequencies } = this.calculateFrequencies(keywords);
const scores = this.calculateScore(termFrequencies, docFrequencies);
// 점수 기준 내림차순 정렬
scores.sort((a, b) =>
b.score !== a.score ? b.score - a.score : b.totalTF - a.totalTF
);
return scores.map(({ id, score, chunkId }) => ({ id, chunkId, score }));
}
/**
* 문서별 단어 빈도 계산
*/
private calculateFrequencies(query: string) {
const pattern = new RegExp(query, "gi");
const termFrequencies: Record<number, Record<string, number>> = {};
const docFrequencies: Record<string, number> = {};
for (const chunk of this.allChunks) {
const text = chunk.text;
const matches = Array.from(text.matchAll(pattern));
const termCounts: Record<string, number> = {};
// 매치된 단어들의 빈도 계산
for (const match of matches) {
const term = match[0].toLowerCase();
termCounts[term] = (termCounts[term] || 0) + 1;
}
if (Object.keys(termCounts).length > 0) {
termFrequencies[chunk.chunkId] = termCounts;
// 문서 빈도 업데이트
for (const term of Object.keys(termCounts)) {
docFrequencies[term] = (docFrequencies[term] || 0) + 1;
}
}
}
return { termFrequencies, docFrequencies };
}
/**
* BM25 점수 계산
*/
private calculateScore(
termFrequencies: Record<number, Record<string, number>>,
docFrequencies: Record<string, number>
): Array<{ id: number; chunkId: number; score: number; totalTF: number }> {
return this.allChunks
.filter((chunk) => termFrequencies[chunk.chunkId])
.map((chunk) => {
const tf = termFrequencies[chunk.chunkId];
const len = chunk.wordCount;
// BM25 점수 계산
const score = Object.keys(tf)
.map((term) => {
const df = docFrequencies[term];
const idf = Math.log((this.N - df + 0.5) / (df + 0.5));
const numerator = tf[term] * (this.k1 + 1);
const denominator =
tf[term] +
this.k1 * (1 - this.b + this.b * (len / this.averageDocLength));
return idf * (numerator / denominator);
})
.reduce((sum, v) => sum + v, 0);
const totalTF = Object.values(tf).reduce((sum, v) => sum + v, 0);
return {
id: chunk.id,
chunkId: chunk.chunkId,
score,
totalTF
};
});
}
}
/**
* 정규식 특수문자 이스케이프
*/
export function escapeRegExp(string: string): string {
return string.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
}
