Mirror/MarginaliaSearch
1
0
Fork 0
mirror of https://github.com/MarginaliaSearch/MarginaliaSearch.git synced 2025-02-24 05:18:58 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

(index) Give ranking components more consistent names

Browse Source
This commit is contained in:
Viktor Lofgren 2024-08-25 10:43:54 +02:00
parent 099133bdbc
commit 4372c8c835
1 changed files with 18 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
code/index/java/nu/marginalia/index/results/IndexResultScoreCalculator.java
View File

@ -311,16 +311,16 @@ public class IndexResultScoreCalculator {
+ temporalBias + temporalBias
+ flagsPenalty; + flagsPenalty;
double tcfAvgDist = rankingParams.tcfAvgDist * (1.0 / calculateAvgMinDistance(positionsQuery, ctx)); double score_avg_dist = rankingParams.tcfAvgDist * (1.0 / calculateAvgMinDistance(positionsQuery, ctx));
double tcfFirstPosition = rankingParams.tcfFirstPosition * (1.0 / Math.sqrt(firstPosition)); double score_firstPosition = rankingParams.tcfFirstPosition * (1.0 / Math.sqrt(firstPosition));
double tcfProximity = rankingParams.tcfProximity * keywordMinDistFac;
double tcfVerbatim = rankingParams.tcfVerbatim * verbatimMatchScore;
double bM25 = rankingParams.bm25Weight * wordFlagsQuery.root.visit(new Bm25GraphVisitor(rankingParams.bm25Params, weightedCounts, length, ctx)); double score_bM25 = rankingParams.bm25Weight * wordFlagsQuery.root.visit(new Bm25GraphVisitor(rankingParams.bm25Params, weightedCounts, length, ctx));
double bFlags = rankingParams.bm25Weight * wordFlagsQuery.root.visit(new TermFlagsGraphVisitor(rankingParams.bm25Params, wordFlagsQuery.data, weightedCounts, ctx)); double score_bFlags = rankingParams.bm25Weight * wordFlagsQuery.root.visit(new TermFlagsGraphVisitor(rankingParams.bm25Params, wordFlagsQuery.data, weightedCounts, ctx));
double score_verbatim = rankingParams.tcfVerbatim * verbatimMatchScore;
double score_proximity = rankingParams.tcfProximity * keywordMinDistFac;
bM25 *= 1.0 / (Math.sqrt(weightedCounts.length + 1)); score_bM25 *= 1.0 / (Math.sqrt(weightedCounts.length + 1));
bFlags *= 1.0 / (Math.sqrt(weightedCounts.length + 1)); score_bFlags *= 1.0 / (Math.sqrt(weightedCounts.length + 1));
if (rankingFactors != null) { if (rankingFactors != null) {
rankingFactors.addDocumentFactor("overall.averageSentenceLengthPenalty", Double.toString(averageSentenceLengthPenalty)); rankingFactors.addDocumentFactor("overall.averageSentenceLengthPenalty", Double.toString(averageSentenceLengthPenalty));
@ -330,14 +330,15 @@ public class IndexResultScoreCalculator {
rankingFactors.addDocumentFactor("overall.topologyBonus", Double.toString(topologyBonus)); rankingFactors.addDocumentFactor("overall.topologyBonus", Double.toString(topologyBonus));
rankingFactors.addDocumentFactor("overall.temporalBias", Double.toString(temporalBias)); rankingFactors.addDocumentFactor("overall.temporalBias", Double.toString(temporalBias));
rankingFactors.addDocumentFactor("overall.flagsPenalty", Double.toString(flagsPenalty)); rankingFactors.addDocumentFactor("overall.flagsPenalty", Double.toString(flagsPenalty));
rankingFactors.addDocumentFactor("overall.verbatimMatchScore", Double.toString(verbatimMatchScore));
rankingFactors.addDocumentFactor("overall.keywordMinDistFac", Double.toString(keywordMinDistFac));
rankingFactors.addDocumentFactor("tcf.avgDist", Double.toString(tcfAvgDist));
rankingFactors.addDocumentFactor("tcf.firstPosition", Double.toString(tcfFirstPosition));
rankingFactors.addDocumentFactor("bm25.main", Double.toString(bM25));
rankingFactors.addDocumentFactor("bm25.flags", Double.toString(bFlags)); rankingFactors.addDocumentFactor("score.bm25-main", Double.toString(score_bM25));
rankingFactors.addDocumentFactor("score.bm25-flags", Double.toString(score_bFlags));
rankingFactors.addDocumentFactor("score.verbatim", Double.toString(score_verbatim));
rankingFactors.addDocumentFactor("score.proximity", Double.toString(score_proximity));
rankingFactors.addDocumentFactor("score.avgDist", Double.toString(score_avg_dist));
rankingFactors.addDocumentFactor("score.firstPosition", Double.toString(score_firstPosition));
rankingFactors.addDocumentFactor("unordered.title", Integer.toString(unorderedMatchInTitleCount)); rankingFactors.addDocumentFactor("unordered.title", Integer.toString(unorderedMatchInTitleCount));
rankingFactors.addDocumentFactor("unordered.heading", Integer.toString(unorderedMatchInHeadingCount)); rankingFactors.addDocumentFactor("unordered.heading", Integer.toString(unorderedMatchInHeadingCount));
@ -379,9 +380,9 @@ public class IndexResultScoreCalculator {
// Renormalize to 0...15, where 0 is the best possible score; // Renormalize to 0...15, where 0 is the best possible score;
// this is a historical artifact of the original ranking function // this is a historical artifact of the original ranking function
double ret = normalize( double ret = normalize(
tcfAvgDist + tcfFirstPosition + tcfProximity + tcfVerbatim score_avg_dist + score_firstPosition + score_proximity + score_verbatim
+ bM25 + score_bM25
+ bFlags + score_bFlags
+ Math.max(0, overallPart), + Math.max(0, overallPart),
-Math.min(0, overallPart)); -Math.min(0, overallPart));