Merge pull request #62 from phenoscape/output-scores

Output score file for regression directly.
phenoscape · May 6, 2020 · 2fb6054 · 2fb6054
2 parents 21b8625 + 7dd5ead
commit 2fb6054
Show file tree

Hide file tree

Showing 2 changed files with 21 additions and 9 deletions.
diff --git a/src/main/scala/org/phenoscape/owl/build/RunPairwiseOWLSim.scala b/src/main/scala/org/phenoscape/owl/build/RunPairwiseOWLSim.scala
@@ -17,7 +17,8 @@ object RunPairwiseOWLSim extends App {
   val ontfile = new File(args(2))
   val profilesFile = new File(args(3))
   val corpus = args(4)
-  val outfile = args(5)
+  val triplesOutfile = args(5)
+  val tsvOutfile = args(6)
 
   val manager = OWLManager.createOWLOntologyManager()
   val ontology = manager.loadOntologyFromOntologyDocument(ontfile)
@@ -39,6 +40,6 @@ object RunPairwiseOWLSim extends App {
   val startIndex = (taskNum - 1) * groupSize
   val group = orderedProfiles.drop(startIndex).take(groupSize)
   println("Computing similarity matrix")
-  owlSim.computeAllSimilarityToCorpusDirectOutput(group.toSet, new File(outfile))
+  owlSim.computeAllSimilarityToCorpusDirectOutput(group.toSet, new File(triplesOutfile), new File(tsvOutfile))
   println("Done: " + new Date())
 }
diff --git a/src/main/scala/org/phenoscape/owl/sim/OWLsim.scala b/src/main/scala/org/phenoscape/owl/sim/OWLsim.scala
@@ -96,27 +96,37 @@ class OWLsim(ontology: OWLOntology, inCorpus: OWLNamedIndividual => Boolean) {
   def computeAllSimilarityToCorpus(inputs: Set[OWLNamedIndividual]): Set[Statement] = (for {
     inputProfile <- inputs.toParArray
     corpusProfile <- individualsInCorpus.toParArray
-    triple <- groupWiseSimilarity(inputProfile, corpusProfile).toTriples
+    (_, triples) = groupWiseSimilarity(inputProfile, corpusProfile).toTriples
+    triple <- triples
   } yield triple).toSet.seq
 
-  def computeAllSimilarityToCorpusDirectOutput(inputs: Set[OWLNamedIndividual], outfile: File): Unit = {
+  def computeAllSimilarityToCorpusDirectOutput(inputs: Set[OWLNamedIndividual], triplesOutfile: File, tsvOutfile: File): Unit = {
     import monix.execution.Scheduler.Implicits.global
-    val outputStream = new FileOutputStream(outfile)
+    val tsvWriter = new PrintWriter(tsvOutfile, "utf-8")
+    tsvWriter.println("?match\t?score\t?query\t?corpusprofile")
+    val outputStream = new FileOutputStream(triplesOutfile)
     val rdfWriter = StreamRDFWriter.getWriterStream(outputStream, RDFFormat.TURTLE_FLAT)
     rdfWriter.start()
     val comparisons = for {
       inputProfile <- Observable.fromIterable(inputs)
       corpusProfile <- Observable.fromIterable(individualsInCorpus)
     } yield (inputProfile, corpusProfile)
     val processed = comparisons.mapParallelUnordered(Runtime.getRuntime.availableProcessors) { case (inputProfile, corpusProfile) =>
-      Task(groupWiseSimilarity(inputProfile, corpusProfile).toTriples)
+      Task {
+        val similarity = groupWiseSimilarity(inputProfile, corpusProfile)
+        val (comparison, triples) = similarity.toTriples
+        val tsvLine = s"$comparison\t${similarity.score}\t${similarity.queryIndividual.getIRI}\t${similarity.corpusIndividual.getIRI}"
+        (tsvLine, triples)
+      }
     }
-    processed.foreachL { triples =>
+    processed.foreachL { case (tsvLine, triples) =>
       //FIXME wasted conversions here
       triples.foreach(triple => rdfWriter.triple(sesameTripleToJena(triple).asTriple))
+      tsvWriter.println(tsvLine)
     }.runSyncUnsafe(Duration.Inf)
     rdfWriter.finish()
     outputStream.close()
+    tsvWriter.close()
   }
 
   def computeAllSimilarityToCorpusJ(inputs: Set[OWLNamedIndividual]): Map[(OWLNamedIndividual, OWLNamedIndividual), Double] = (for {
@@ -342,7 +352,7 @@ final case class GroupWiseSimilarity(queryIndividual: OWLNamedIndividual, corpus
 
   import GroupWiseSimilarity._
 
-  def toTriples: Set[Statement] = {
+  def toTriples: (URI, Set[Statement]) = {
     val self = new URIImpl(OntUtil.nextIRI.toString)
     val micasTriples = for {
       pair <- pairs
@@ -354,10 +364,11 @@ final case class GroupWiseSimilarity(queryIndividual: OWLNamedIndividual, corpus
       node <- distinctSubsumers
       term <- node.classes
     } yield new StatementImpl(self, has_subsumer, new URIImpl(term.getIRI.toString))
-    Set(
+    val triples = Set(
       new StatementImpl(self, combined_score, new NumericLiteralImpl(score)),
       new StatementImpl(self, for_query_profile, new URIImpl(queryIndividual.getIRI.toString)),
       new StatementImpl(self, for_corpus_profile, new URIImpl(corpusIndividual.getIRI.toString))) ++ subsumerTriples ++ micasTriples
+    self -> triples.toSet
   }
 
 }