nyancrimew
/
LatinIME
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge "Start-of-sentence should include newlines and non-period terminators."

main
Tom Ouyang 2015-03-17 22:24:59 +00:00 committed by Android (Google) Code Review
parent ea727b6b71 5aeb092130
commit edcf5853d1
2 changed files with 32 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
java/src/com/android/inputmethod/latin/utils/NgramContextUtils.java
View File

@ -31,6 +31,7 @@ public final class NgramContextUtils {
// Intentional empty constructor for utility class.
}
private static final Pattern NEWLINE_REGEX = Pattern.compile("[\\r\\n]+");
private static final Pattern SPACE_REGEX = Pattern.compile("\\s+");
// Get context information from nth word before the cursor. n = 1 retrieves the words
// immediately before the cursor, n = 2 retrieves the words before that, and so on. This splits
@ -58,7 +59,11 @@ public final class NgramContextUtils {
public static NgramContext getNgramContextFromNthPreviousWord(final CharSequence prev,
final SpacingAndPunctuations spacingAndPunctuations, final int n) {
if (prev == null) return NgramContext.EMPTY_PREV_WORDS_INFO;
final String[] w = SPACE_REGEX.split(prev);
final String[] lines = NEWLINE_REGEX.split(prev);
if (lines.length == 0) {
return new NgramContext(WordInfo.BEGINNING_OF_SENTENCE_WORD_INFO);
}
final String[] w = SPACE_REGEX.split(lines[lines.length - 1]);
final WordInfo[] prevWordsInfo =
new WordInfo[DecoderSpecificConstants.MAX_PREV_WORD_COUNT_FOR_N_GRAM];
Arrays.fill(prevWordsInfo, WordInfo.EMPTY_WORD_INFO);
@ -81,16 +86,17 @@ public final class NgramContextUtils {
prevWordsInfo[i] = WordInfo.BEGINNING_OF_SENTENCE_WORD_INFO;
break;
}
final String focusedWord = w[focusedWordIndex];
// If the word is, the context is beginning-of-sentence.
// If the word is empty, the context is beginning-of-sentence.
final int length = focusedWord.length();
if (length <= 0) {
prevWordsInfo[i] = WordInfo.BEGINNING_OF_SENTENCE_WORD_INFO;
break;
}
// If ends in a sentence separator, the context is beginning-of-sentence.
// If the word ends in a sentence terminator, the context is beginning-of-sentence.
final char lastChar = focusedWord.charAt(length - 1);
if (spacingAndPunctuations.isSentenceSeparator(lastChar)) {
if (spacingAndPunctuations.isSentenceTerminator(lastChar)) {
prevWordsInfo[i] = WordInfo.BEGINNING_OF_SENTENCE_WORD_INFO;
break;
}

22
tests/src/com/android/inputmethod/latin/NgramContextTests.java
View File

@ -17,6 +17,8 @@
package com.android.inputmethod.latin;
import com.android.inputmethod.latin.NgramContext.WordInfo;
import com.android.inputmethod.latin.settings.SpacingAndPunctuations;
import com.android.inputmethod.latin.utils.NgramContextUtils;
import android.test.AndroidTestCase;
import android.test.suitebuilder.annotation.SmallTest;
@ -120,4 +122,24 @@ public class NgramContextTests extends AndroidTestCase {
assertEquals(1, ngramContext_a_empty.extractPrevWordsContextArray().length);
assertEquals("a", ngramContext_a_empty.extractPrevWordsContextArray()[0]);
}
public void testGetNgramContextFromNthPreviousWord() {
SpacingAndPunctuations spacingAndPunctuations = new SpacingAndPunctuations(
mContext.getResources());
assertEquals("<S>", NgramContextUtils.getNgramContextFromNthPreviousWord("",
spacingAndPunctuations, 1).extractPrevWordsContext());
assertEquals("<S> b", NgramContextUtils.getNgramContextFromNthPreviousWord("a. b ",
spacingAndPunctuations, 1).extractPrevWordsContext());
assertEquals("<S> b", NgramContextUtils.getNgramContextFromNthPreviousWord("a? b ",
spacingAndPunctuations, 1).extractPrevWordsContext());
assertEquals("<S> b", NgramContextUtils.getNgramContextFromNthPreviousWord("a! b ",
spacingAndPunctuations, 1).extractPrevWordsContext());
assertEquals("<S> b", NgramContextUtils.getNgramContextFromNthPreviousWord("a\nb ",
spacingAndPunctuations, 1).extractPrevWordsContext());
assertEquals("<S> a b", NgramContextUtils.getNgramContextFromNthPreviousWord("a b ",
spacingAndPunctuations, 1).extractPrevWordsContext());
assertFalse(NgramContextUtils
.getNgramContextFromNthPreviousWord("a b c d e", spacingAndPunctuations, 1)
.extractPrevWordsContext().startsWith("<S>"));
}
}