added qljm and qld options for Jelinek-Mercer and Dirichlet QL scoring (

#491)
castorini · Nov 30, 2018 · c1aac5e · c1aac5e
1 parent e5b87f0
commit c1aac5e
Show file tree

Hide file tree

Showing 2 changed files with 14 additions and 1 deletion.
diff --git a/src/main/java/io/anserini/search/SearchArgs.java b/src/main/java/io/anserini/search/SearchArgs.java
@@ -103,6 +103,15 @@ public class SearchArgs {
    * setting seems reasonable and does not contradict Zhai and Lafferty.
    */
 
+  @Option(name = "-qld", usage = "use query likelihood Dirichlet scoring model")
+  public boolean qld = false;
+
+  @Option(name = "-qljm", usage = "use query likelihood Jelinek Mercer scoring model")
+  public boolean qljm = false;
+
+  @Option(name = "-qljm.lambda", handler = StringArrayOptionHandler.class, usage = "Jelinek Mercer smoothing parameter")
+  public String[] qljm_lambda = new String[] {"0.1"};
+
   @Option(name = "-bm25", usage = "use BM25 scoring model")
   public boolean bm25 = false;
 

diff --git a/src/main/java/io/anserini/search/SearchCollection.java b/src/main/java/io/anserini/search/SearchCollection.java
@@ -215,10 +215,14 @@ public void close() throws IOException {
   public List<TaggedSimilarity> constructSimiliries() {
     // Figure out which scoring model to use.
     List<TaggedSimilarity> similarities = new ArrayList<>();
-    if (args.ql) {
+    if (args.ql || args.qld) {
       for (String mu : args.mu) {
         similarities.add(new TaggedSimilarity(new LMDirichletSimilarity(Float.valueOf(mu)), "mu:"+mu));
       }
+    } else if (args.qljm) {
+      for (String lambda : args.qljm_lambda) {
+        similarities.add(new TaggedSimilarity(new LMJelinekMercerSimilarity(Float.valueOf(lambda)), "lambda:" + lambda));
+      }
     } else if (args.bm25) {
       for (String k1 : args.k1) {
         for (String b : args.b) {