1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
|
/*
* Copyright (C) 2014 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.android.inputmethod.latin.utils;
import android.util.Log;
import com.android.inputmethod.latin.Dictionary;
import com.android.inputmethod.latin.DictionaryFacilitatorForSuggest;
import com.android.inputmethod.latin.PrevWordsInfo;
import com.android.inputmethod.latin.settings.SpacingAndPunctuations;
import java.util.ArrayList;
import java.util.Locale;
// Note: this class is used as a parameter type of a native method. You should be careful when you
// rename this class or field name. See BinaryDictionary#addMultipleDictionaryEntriesNative().
public final class LanguageModelParam {
private static final String TAG = LanguageModelParam.class.getSimpleName();
private static final boolean DEBUG = false;
private static final boolean DEBUG_TOKEN = false;
// For now, these probability values are being referred to only when we add new entries to
// decaying dynamic binary dictionaries. When these are referred to, what matters is 0 or
// non-0. Thus, it's not meaningful to compare 10, 100, and so on.
// TODO: Revise the logic in ForgettingCurveUtils in native code.
private static final int UNIGRAM_PROBABILITY_FOR_VALID_WORD = 100;
private static final int UNIGRAM_PROBABILITY_FOR_OOV_WORD = Dictionary.NOT_A_PROBABILITY;
private static final int BIGRAM_PROBABILITY_FOR_VALID_WORD = 10;
private static final int BIGRAM_PROBABILITY_FOR_OOV_WORD = Dictionary.NOT_A_PROBABILITY;
public final String mTargetWord;
public final int[] mWord0;
public final int[] mWord1;
// TODO: this needs to be a list of shortcuts
public final int[] mShortcutTarget;
public final int mUnigramProbability;
public final int mBigramProbability;
public final int mShortcutProbability;
public final boolean mIsNotAWord;
public final boolean mIsBlacklisted;
// Time stamp in seconds.
public final int mTimestamp;
// Constructor for unigram. TODO: support shortcuts
public LanguageModelParam(final String word, final int unigramProbability,
final int timestamp) {
this(null /* word0 */, word, unigramProbability, Dictionary.NOT_A_PROBABILITY, timestamp);
}
// Constructor for unigram and bigram.
public LanguageModelParam(final String word0, final String word1,
final int unigramProbability, final int bigramProbability,
final int timestamp) {
mTargetWord = word1;
mWord0 = (word0 == null) ? null : StringUtils.toCodePointArray(word0);
mWord1 = StringUtils.toCodePointArray(word1);
mShortcutTarget = null;
mUnigramProbability = unigramProbability;
mBigramProbability = bigramProbability;
mShortcutProbability = Dictionary.NOT_A_PROBABILITY;
mIsNotAWord = false;
mIsBlacklisted = false;
mTimestamp = timestamp;
}
// Process a list of words and return a list of {@link LanguageModelParam} objects.
public static ArrayList<LanguageModelParam> createLanguageModelParamsFrom(
final ArrayList<String> tokens, final int timestamp,
final DictionaryFacilitatorForSuggest dictionaryFacilitator,
final SpacingAndPunctuations spacingAndPunctuations,
final DistracterFilter distracterFilter) {
final ArrayList<LanguageModelParam> languageModelParams =
CollectionUtils.newArrayList();
final int N = tokens.size();
PrevWordsInfo prevWordsInfo = new PrevWordsInfo(null);
for (int i = 0; i < N; ++i) {
final String tempWord = tokens.get(i);
if (StringUtils.isEmptyStringOrWhiteSpaces(tempWord)) {
// just skip this token
if (DEBUG_TOKEN) {
Log.d(TAG, "--- isEmptyStringOrWhiteSpaces: \"" + tempWord + "\"");
}
continue;
}
if (!DictionaryInfoUtils.looksValidForDictionaryInsertion(
tempWord, spacingAndPunctuations)) {
if (DEBUG_TOKEN) {
Log.d(TAG, "--- not looksValidForDictionaryInsertion: \""
+ tempWord + "\"");
}
// Sentence terminator found. Split.
prevWordsInfo = new PrevWordsInfo(null);
continue;
}
if (DEBUG_TOKEN) {
Log.d(TAG, "--- word: \"" + tempWord + "\"");
}
final LanguageModelParam languageModelParam =
detectWhetherVaildWordOrNotAndGetLanguageModelParam(
prevWordsInfo, tempWord, timestamp, dictionaryFacilitator,
distracterFilter);
if (languageModelParam == null) {
continue;
}
languageModelParams.add(languageModelParam);
prevWordsInfo = new PrevWordsInfo(languageModelParam.mTargetWord);
}
return languageModelParams;
}
private static LanguageModelParam detectWhetherVaildWordOrNotAndGetLanguageModelParam(
final PrevWordsInfo prevWordsInfo, final String targetWord, final int timestamp,
final DictionaryFacilitatorForSuggest dictionaryFacilitator,
final DistracterFilter distracterFilter) {
final Locale locale = dictionaryFacilitator.getLocale();
if (locale == null) {
return null;
}
// TODO: Though targetWord is an IV (in-vocabulary) word, we should still apply
// distracterFilter in the following code. If targetWord is a distracter,
// it should be filtered out.
if (dictionaryFacilitator.isValidWord(targetWord, false /* ignoreCase */)) {
return createAndGetLanguageModelParamOfWord(prevWordsInfo, targetWord, timestamp,
true /* isValidWord */, locale);
}
final String lowerCaseTargetWord = targetWord.toLowerCase(locale);
if (dictionaryFacilitator.isValidWord(lowerCaseTargetWord, false /* ignoreCase */)) {
// Add the lower-cased word.
return createAndGetLanguageModelParamOfWord(prevWordsInfo, lowerCaseTargetWord,
timestamp, true /* isValidWord */, locale);
}
// Treat the word as an OOV word. The following statement checks whether this OOV
// is a distracter to words in dictionaries. Being a distracter means the OOV word is
// too close to a common word in dictionaries (e.g., the OOV "mot" is very close to "not").
// Adding such a word to dictonaries would interfere with entering in-dictionary words. For
// example, adding "mot" to dictionaries might interfere with entering "not".
// This kind of OOV should be filtered out.
if (distracterFilter.isDistracterToWordsInDictionaries(prevWordsInfo, targetWord, locale)) {
return null;
}
return createAndGetLanguageModelParamOfWord(prevWordsInfo, targetWord, timestamp,
false /* isValidWord */, locale);
}
private static LanguageModelParam createAndGetLanguageModelParamOfWord(
final PrevWordsInfo prevWordsInfo, final String targetWord, final int timestamp,
final boolean isValidWord, final Locale locale) {
final String word;
if (StringUtils.getCapitalizationType(targetWord) == StringUtils.CAPITALIZE_FIRST
&& prevWordsInfo.mPrevWord == null && !isValidWord) {
word = targetWord.toLowerCase(locale);
} else {
word = targetWord;
}
final int unigramProbability = isValidWord ?
UNIGRAM_PROBABILITY_FOR_VALID_WORD : UNIGRAM_PROBABILITY_FOR_OOV_WORD;
if (prevWordsInfo.mPrevWord == null) {
if (DEBUG) {
Log.d(TAG, "--- add unigram: current("
+ (isValidWord ? "Valid" : "OOV") + ") = " + word);
}
return new LanguageModelParam(word, unigramProbability, timestamp);
}
if (DEBUG) {
Log.d(TAG, "--- add bigram: prev = " + prevWordsInfo.mPrevWord + ", current("
+ (isValidWord ? "Valid" : "OOV") + ") = " + word);
}
final int bigramProbability = isValidWord ?
BIGRAM_PROBABILITY_FOR_VALID_WORD : BIGRAM_PROBABILITY_FOR_OOV_WORD;
return new LanguageModelParam(prevWordsInfo.mPrevWord, word, unigramProbability,
bigramProbability, timestamp);
}
}
|
