Rename combine to fuse

Author: Joseph Smith

media-api/app/lib/elasticsearch/ElasticSearch.scala

@@ -261,7 +261,7 @@ class ElasticSearch(
     vecWeight: Double,
     filterOpt: Option[Query]
   )(implicit ex: ExecutionContext, logMarker: LogMarker): Future[SearchResults] = {
-    import HybridResult.{resolveHitAndFillInSemanticScore, combineScoresAndGetTopK}
+    import HybridResult.{resolveHitAndFillInSemanticScore, fuseScoresAndGetTopK}
 
     val lexicalQuery = createMultiMatchQuery(query, operator = Or)
     val lexicalSearchRequest = lexicalRequest(lexicalQuery, k, filterOpt)
@@ -302,7 +302,7 @@ class ElasticSearch(
       val resultsWithSemanticScoresFilledIn = distinctHits.flatMap { hit =>
         resolveHitAndFillInSemanticScore(hit, queryEmbedding, resolveHit)
       }
-      val topK = combineScoresAndGetTopK(resultsWithSemanticScoresFilledIn, vecWeight, k)
+      val topK = fuseScoresAndGetTopK(resultsWithSemanticScoresFilledIn, vecWeight, k)
       topK.foreach { r =>
         logger.info(logMarker, s"hybrid result ${r.id}: lexicalScore=${r.lexicalScore} semanticScore=${r.semanticScore} " +
           s"originallyFromLexical=${lexicalIds.contains(r.id)} originallyFromSemantic=${semanticIds.contains(r.id)}")

media-api/app/lib/elasticsearch/HybridResult.scala

@@ -38,7 +38,7 @@ object HybridResult {
     if (max == theoreticalMin) 0.0
     else (score - theoreticalMin) / (max - theoreticalMin)
 
-  def combinedScore(
+  def fuseScores(
     result: HybridResult,
     maxLexicalScore: Double,
     maxSemanticScore: Double,
@@ -51,7 +51,7 @@ object HybridResult {
     (vecWeight * normedSemanticScore) + ((1 - vecWeight) * normedLexicalScore)
   }
 
-  def combineScoresAndGetTopK(
+  def fuseScoresAndGetTopK(
     results: List[HybridResult],
     vecWeight: Double,
     k: Int
@@ -68,7 +68,7 @@ object HybridResult {
       maxSemanticScore <- results.map(_.semanticScore).maxOption
     } yield {
       results
-        .sortBy(combinedScore(_, maxLexicalScore, maxSemanticScore, vecWeight))(Ordering[Double].reverse)
+        .sortBy(fuseScores(_, maxLexicalScore, maxSemanticScore, vecWeight))(Ordering[Double].reverse)
         .take(k)
     }).getOrElse(List())
   }

media-api/test/lib/elasticsearch/HybridResultTest.scala

@@ -10,7 +10,7 @@ import play.api.libs.json.Json
 
 class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with Tolerance with Fixtures {
 
-  import HybridResult.{combineScoresAndGetTopK, combinedScore, normalise, resolveHitAndFillInSemanticScore}
+  import HybridResult.{fuseScoresAndGetTopK, fuseScores, normalise, resolveHitAndFillInSemanticScore}
 
   private val tolerance = 1e-9
 
@@ -153,7 +153,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
     it("blends the normalised lexical and semantic scores using vecWeight") {
       // normedLexical = 2/4 = 0.5, normedSemantic = (0 + 1)/(1 + 1) = 0.5
       // combined = 0.25 * 0.5 + 0.75 * 0.5 = 0.5
-      val score = combinedScore(
+      val score = fuseScores(
         hybridResult("blend", lexicalScore = 2.0, semanticScore = 0.0),
         maxLexicalScore = 4.0,
         maxSemanticScore = 1.0,
@@ -163,7 +163,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
     }
 
     it("uses only the lexical score when vecWeight is 0") {
-      val score = combinedScore(
+      val score = fuseScores(
         hybridResult("lexical-only", lexicalScore = 2.0, semanticScore = 1.0),
         maxLexicalScore = 4.0,
         maxSemanticScore = 1.0,
@@ -173,7 +173,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
     }
 
     it("uses only the semantic score when vecWeight is 1") {
-      val score = combinedScore(
+      val score = fuseScores(
         hybridResult("semantic-only", lexicalScore = 2.0, semanticScore = 0.0),
         maxLexicalScore = 4.0,
         maxSemanticScore = 1.0,
@@ -184,7 +184,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
     }
 
     it("contributes 0 from the lexical side when the max lexical score is 0") {
-      val score = combinedScore(
+      val score = fuseScores(
         hybridResult("no-lexical", lexicalScore = 0.0, semanticScore = 1.0),
         maxLexicalScore = 0.0,
         maxSemanticScore = 1.0,
@@ -196,7 +196,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
     }
 
     it("contributes 0 from the semantic side when the max semantic score is -1") {
-      val score = combinedScore(
+      val score = fuseScores(
         hybridResult("no-semantic", lexicalScore = 4.0, semanticScore = -1.0),
         maxLexicalScore = 4.0,
         maxSemanticScore = -1.0,
@@ -210,7 +210,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
 
   describe("combineScoresAndGetTopK") {
     it("returns an empty list if the input is empty") {
-      combineScoresAndGetTopK(List(), vecWeight = 0.0, k = 3) should be (List())
+      fuseScoresAndGetTopK(List(), vecWeight = 0.0, k = 3) should be (List())
     }
 
     it("doesn't break when tha max lexical score is 0") {
@@ -228,7 +228,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
         hybridResult("mid-lexical", lexicalScore = 5.0, semanticScore = 0.0)
       )
 
-      val ranked = combineScoresAndGetTopK(results, vecWeight = 0.0, k = 3)
+      val ranked = fuseScoresAndGetTopK(results, vecWeight = 0.0, k = 3)
 
       ranked.map(_.id) should be(List("high-lexical", "mid-lexical", "low-lexical"))
     }
@@ -240,7 +240,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
         hybridResult("mid-lexical", lexicalScore = 5.0, semanticScore = -1.0)
       )
 
-      val ranked = combineScoresAndGetTopK(results, vecWeight = 0.5, k = 3)
+      val ranked = fuseScoresAndGetTopK(results, vecWeight = 0.5, k = 3)
 
       ranked.map(_.id) should be(List("high-lexical", "mid-lexical", "low-lexical"))
     }
@@ -252,7 +252,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
         hybridResult("mid-semantic", lexicalScore = 5.0, semanticScore = 0.0)
       )
 
-      val ranked = combineScoresAndGetTopK(results, vecWeight = 1.0, k = 3)
+      val ranked = fuseScoresAndGetTopK(results, vecWeight = 1.0, k = 3)
 
       ranked.map(_.id) should be(List("high-semantic", "mid-semantic", "low-semantic"))
     }
@@ -264,7 +264,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
         hybridResult("mid-semantic", lexicalScore = 0.0, semanticScore = 0.0)
       )
 
-      val ranked = combineScoresAndGetTopK(results, vecWeight = 0.5, k = 3)
+      val ranked = fuseScoresAndGetTopK(results, vecWeight = 0.5, k = 3)
 
       ranked.map(_.id) should be(List("high-semantic", "mid-semantic", "low-semantic"))
     }
@@ -276,7 +276,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
         hybridResult("c", lexicalScore = 5.0, semanticScore = 1.0)
       )
 
-      val ranked = combineScoresAndGetTopK(results, vecWeight = 0.0, k = 2)
+      val ranked = fuseScoresAndGetTopK(results, vecWeight = 0.0, k = 2)
 
       ranked should have size 2
       ranked.map(_.id) should be(List("b", "c"))
@@ -292,7 +292,7 @@ class HybridResultTest extends AnyFunSpec with Matchers with OptionValues with T
         hybridResult("blend", lexicalScore = 2.0, semanticScore = 0.0)
       )
 
-      val ranked = combineScoresAndGetTopK(results, vecWeight = 0.25, k = 3)
+      val ranked = fuseScoresAndGetTopK(results, vecWeight = 0.25, k = 3)
 
       val blend = ranked.find(_.id == "blend").value
       // Recompute the expected combined score independently of ordering.