Don't remove indexable symbols when stemming

1 files changed, 7 insertions, 1 deletions

diff --git a/linguistics/src/main/java/com/yahoo/language/simple/SimpleTokenizer.java b/linguistics/src/main/java/com/yahoo/language/simple/SimpleTokenizer.java
index 2728249333e..d86ca30a632 100644
--- a/linguistics/src/main/java/com/yahoo/language/simple/SimpleTokenizer.java
+++ b/linguistics/src/main/java/com/yahoo/language/simple/SimpleTokenizer.java
@@ -67,7 +67,7 @@ public class SimpleTokenizer implements Tokenizer {
         for (int prev = 0, next = Character.charCount(nextCode); next <= input.length(); ) {
             nextCode = next < input.length() ? input.codePointAt(next) : SPACE_CODE;
             TokenType nextType = SimpleTokenType.valueOf(nextCode);
-            if (!prevType.isIndexable() || !nextType.isIndexable()) {
+            if (isAtTokenBoundary(prevType, nextType)) {
                 String original = input.substring(prev, next);
                 tokens.add(new SimpleToken(original).setOffset(prev)
                                                     .setType(tokenType)
@@ -84,6 +84,12 @@ public class SimpleTokenizer implements Tokenizer {
         return tokens;
     }
 
+    private boolean isAtTokenBoundary(TokenType prevType, TokenType nextType) {
+        // Always index each symbol as a token
+        if (prevType == TokenType.INDEXABLE_SYMBOL || nextType == TokenType.INDEXABLE_SYMBOL) return true;
+        return !prevType.isIndexable() || !nextType.isIndexable();
+    }
+
     private TokenType determineType(TokenType tokenType, TokenType characterType) {
         if (characterType == TokenType.ALPHABETIC) return TokenType.ALPHABETIC;
         return tokenType;