diff options
Diffstat (limited to 'opennlp-linguistics/src/test/java/com/yahoo/language/opennlp/OpenNlpTokenizationTestCase.java')
-rw-r--r-- | opennlp-linguistics/src/test/java/com/yahoo/language/opennlp/OpenNlpTokenizationTestCase.java | 19 |
1 files changed, 17 insertions, 2 deletions
diff --git a/opennlp-linguistics/src/test/java/com/yahoo/language/opennlp/OpenNlpTokenizationTestCase.java b/opennlp-linguistics/src/test/java/com/yahoo/language/opennlp/OpenNlpTokenizationTestCase.java index a5daf7f0531..78412f94fd4 100644 --- a/opennlp-linguistics/src/test/java/com/yahoo/language/opennlp/OpenNlpTokenizationTestCase.java +++ b/opennlp-linguistics/src/test/java/com/yahoo/language/opennlp/OpenNlpTokenizationTestCase.java @@ -150,8 +150,7 @@ public class OpenNlpTokenizationTestCase { @Test public void testIndexability() { String input = "tafsirnya\u0648\u0643\u064F\u0646\u0652"; - for (StemMode stemMode : new StemMode[] { StemMode.NONE, - StemMode.SHORTEST }) { + for (StemMode stemMode : new StemMode[] { StemMode.NONE, StemMode.SHORTEST }) { for (Language l : List.of(Language.INDONESIAN, Language.ENGLISH, Language.ARABIC)) { for (boolean accentDrop : new boolean[] { true, false }) { for (Token token : tokenizer.tokenize(input, l, stemMode, accentDrop)) { @@ -165,6 +164,22 @@ public class OpenNlpTokenizationTestCase { } @Test + public void testTokenizeEmojis() { + String emoji1 = "\uD83D\uDD2A"; // 🔪 + Iterator<Token> tokens1 = tokenizer.tokenize(emoji1, Language.ENGLISH, StemMode.ALL, true).iterator(); + assertTrue(tokens1.hasNext()); + assertEquals(emoji1, tokens1.next().getTokenString()); + assertFalse(tokens1.hasNext()); + + String emoji2 = "\uD83D\uDE00"; // 😀 + Iterator<Token> tokens2 = tokenizer.tokenize(emoji1 + emoji2, Language.ENGLISH, StemMode.ALL, true).iterator(); + assertTrue(tokens2.hasNext()); + assertEquals(emoji1, tokens2.next().getTokenString()); + assertEquals(emoji2, tokens2.next().getTokenString()); + assertFalse(tokens2.hasNext()); + } + + @Test public void testTokenTypes() { testTokenTypes(Language.ENGLISH); testTokenTypes(Language.SPANISH); |