Replace UrlcharSequenceNormalizer with one with an improved regex

5 files changed, 80 insertions, 6 deletions

diff --git a/linguistics-components/src/main/java/com/yahoo/language/opennlp/LanguageDetectorFactory.java b/linguistics-components/src/main/java/com/yahoo/language/opennlp/LanguageDetectorFactory.java
new file mode 100644
index 00000000000..aa4387bcc45
--- /dev/null
+++ b/linguistics-components/src/main/java/com/yahoo/language/opennlp/LanguageDetectorFactory.java
@@ -0,0 +1,29 @@
+// Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+package com.yahoo.language.opennlp;
+
+import opennlp.tools.langdetect.DefaultLanguageDetectorContextGenerator;
+import opennlp.tools.langdetect.LanguageDetectorContextGenerator;
+import opennlp.tools.util.normalizer.EmojiCharSequenceNormalizer;
+import opennlp.tools.util.normalizer.NumberCharSequenceNormalizer;
+import opennlp.tools.util.normalizer.ShrinkCharSequenceNormalizer;
+import opennlp.tools.util.normalizer.TwitterCharSequenceNormalizer;
+
+/**
+ * Overrides the UrlCharSequenceNormalizer, which has a bad regex, until fixed: https://issues.apache.org/jira/browse/OPENNLP-1350
+ *
+ * @author jonmv
+ */
+@SuppressWarnings("unused") // Loaded by black magic.
+public class LanguageDetectorFactory extends opennlp.tools.langdetect.LanguageDetectorFactory {
+
+    @Override
+    public LanguageDetectorContextGenerator getContextGenerator() {
+        return new DefaultLanguageDetectorContextGenerator(1, 3,
+                                                           EmojiCharSequenceNormalizer.getInstance(),
+                                                           UrlCharSequenceNormalizer.getInstance(),
+                                                           TwitterCharSequenceNormalizer.getInstance(),
+                                                           NumberCharSequenceNormalizer.getInstance(),
+                                                           ShrinkCharSequenceNormalizer.getInstance());
+    }
+
+}
diff --git a/linguistics-components/src/main/java/com/yahoo/language/opennlp/UrlCharSequenceNormalizer.java b/linguistics-components/src/main/java/com/yahoo/language/opennlp/UrlCharSequenceNormalizer.java
new file mode 100644
index 00000000000..883319e2f8b
--- /dev/null
+++ b/linguistics-components/src/main/java/com/yahoo/language/opennlp/UrlCharSequenceNormalizer.java
@@ -0,0 +1,31 @@
+// Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+package com.yahoo.language.opennlp;
+
+import opennlp.tools.util.normalizer.CharSequenceNormalizer;
+
+import java.util.regex.Pattern;
+
+/**
+ * Modifies {@link opennlp.tools.util.normalizer.UrlCharSequenceNormalizer} to avoid the bad email regex.
+ *
+ * @author jonmv
+ */
+public class UrlCharSequenceNormalizer implements CharSequenceNormalizer {
+
+    private static final Pattern URL_REGEX =
+            Pattern.compile("https?://[-_.?&~;+=/#0-9A-Za-z]+");
+    private static final Pattern MAIL_REGEX =
+            Pattern.compile("(?<![-+_.0-9A-Za-z])[-+_.0-9A-Za-z]+@[-0-9A-Za-z]+[-.0-9A-Za-z]+");
+
+    private static final UrlCharSequenceNormalizer INSTANCE = new UrlCharSequenceNormalizer();
+
+    public static UrlCharSequenceNormalizer getInstance() {
+        return INSTANCE;
+    }
+
+    public CharSequence normalize(CharSequence text) {
+        String modified = URL_REGEX.matcher(text).replaceAll(" ");
+        return MAIL_REGEX.matcher(modified).replaceAll(" ");
+    }
+
+}
diff --git a/linguistics-components/src/main/resources/models/langdetect-183.bin b/linguistics-components/src/main/resources/models/langdetect-183.bin
index 0b4ea89690e..c3cde217050 100644
--- a/linguistics-components/src/main/resources/models/langdetect-183.bin
+++ b/linguistics-components/src/main/resources/models/langdetect-183.bin
diff --git a/linguistics-components/src/test/java/com/yahoo/language/opennlp/UrlCharSequenceNormalizerTest.java b/linguistics-components/src/test/java/com/yahoo/language/opennlp/UrlCharSequenceNormalizerTest.java
new file mode 100644
index 00000000000..a8c637bc6ec
--- /dev/null
+++ b/linguistics-components/src/test/java/com/yahoo/language/opennlp/UrlCharSequenceNormalizerTest.java
@@ -0,0 +1,20 @@
+// Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+package com.yahoo.language.opennlp;
+
+import org.junit.Test;
+
+import static org.junit.Assert.assertEquals;
+
+/**
+ * @author jonmv
+ */
+public class UrlCharSequenceNormalizerTest {
+
+    @Test
+    public void testNormalization() {
+        String text = "xxx+yyy_.dude@mail.com foo bar@baz_bax https://host.tld/path?query=boo a@b §boo@boo";
+        assertEquals("  foo  _bax   a@b § ",
+                     UrlCharSequenceNormalizer.getInstance().normalize(text));
+    }
+
+}
diff --git a/linguistics/src/main/resources/configdefinitions/language.opennlp.opennlp-linguistics.def b/linguistics/src/main/resources/configdefinitions/language.opennlp.opennlp-linguistics.def
deleted file mode 100644
index 361a8a5f50c..00000000000
--- a/linguistics/src/main/resources/configdefinitions/language.opennlp.opennlp-linguistics.def
+++ /dev/null
@@ -1,6 +0,0 @@
-# Copyright Yahoo. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
-namespace=language.opennlp
-
-# Enable Optimaize language detector
-detector.enableOptimaize bool  default=true
-