diff --git a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
index e79fb71..28aaf2d 100644
--- a/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
+++ b/native/jni/com_android_inputmethod_latin_BinaryDictionary.cpp
@@ -28,6 +28,7 @@
 #include "suggest/core/dictionary/property/unigram_property.h"
 #include "suggest/core/dictionary/property/word_property.h"
 #include "suggest/core/result/suggestion_results.h"
+#include "suggest/core/session/prev_words_info.h"
 #include "suggest/core/suggest_options.h"
 #include "suggest/policyimpl/dictionary/structure/dictionary_structure_with_buffer_policy_factory.h"
 #include "utils/char_utils.h"
@@ -247,15 +248,15 @@
     float languageWeight;
     env->GetFloatArrayRegion(inOutLanguageWeight, 0, 1 /* len */, &languageWeight);
     SuggestionResults suggestionResults(MAX_RESULTS);
+    const PrevWordsInfo prevWordsInfo(prevWordCodePoints, prevWordCodePointsLength,
+            false /* isStartOfSentence */);
     if (givenSuggestOptions.isGesture() || inputSize > 0) {
         // TODO: Use SuggestionResults to return suggestions.
         dictionary->getSuggestions(pInfo, traverseSession, xCoordinates, yCoordinates,
-                times, pointerIds, inputCodePoints, inputSize, prevWordCodePoints,
-                prevWordCodePointsLength, &givenSuggestOptions, languageWeight,
-                &suggestionResults);
+                times, pointerIds, inputCodePoints, inputSize, &prevWordsInfo,
+                &givenSuggestOptions, languageWeight, &suggestionResults);
     } else {
-        dictionary->getPredictions(prevWordCodePoints, prevWordCodePointsLength,
-                &suggestionResults);
+        dictionary->getPredictions(&prevWordsInfo, &suggestionResults);
     }
     suggestionResults.outputSuggestions(env, outSuggestionCount, outCodePointsArray,
             outScoresArray, outSpaceIndicesArray, outTypesArray,
@@ -282,8 +283,8 @@
     int word1CodePoints[word1Length];
     env->GetIntArrayRegion(word0, 0, word0Length, word0CodePoints);
     env->GetIntArrayRegion(word1, 0, word1Length, word1CodePoints);
-    return dictionary->getBigramProbability(word0CodePoints, word0Length, word1CodePoints,
-            word1Length);
+    const PrevWordsInfo prevWordsInfo(word0CodePoints, word0Length, false /* isStartOfSentence */);
+    return dictionary->getBigramProbability(&prevWordsInfo, word1CodePoints, word1Length);
 }
 
 // Method to iterate all words in the dictionary for makedict.
diff --git a/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp b/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp
index 5cc20aa..7660641 100644
--- a/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp
+++ b/native/jni/com_android_inputmethod_latin_DicTraverseSession.cpp
@@ -22,6 +22,7 @@
 #include "jni.h"
 #include "jni_common.h"
 #include "suggest/core/session/dic_traverse_session.h"
+#include "suggest/core/session/prev_words_info.h"
 
 namespace latinime {
 class Dictionary;
@@ -39,12 +40,14 @@
     }
     Dictionary *dict = reinterpret_cast<Dictionary *>(dictionary);
     if (!previousWord) {
-        ts->init(dict, 0 /* prevWord */, 0 /* prevWordLength*/, 0 /* suggestOptions */);
+        PrevWordsInfo prevWordsInfo;
+        ts->init(dict, &prevWordsInfo, 0 /* suggestOptions */);
         return;
     }
     int prevWord[previousWordLength];
     env->GetIntArrayRegion(previousWord, 0, previousWordLength, prevWord);
-    ts->init(dict, prevWord, previousWordLength, 0 /* suggestOptions */);
+    PrevWordsInfo prevWordsInfo(prevWord, previousWordLength, false /* isStartOfSentence */);
+    ts->init(dict, &prevWordsInfo, 0 /* suggestOptions */);
 }
 
 static void latinime_releaseDicTraverseSession(JNIEnv *env, jclass clazz, jlong traverseSession) {
diff --git a/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp b/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp
index f793363..847fa1b 100644
--- a/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp
+++ b/native/jni/src/suggest/core/dictionary/bigram_dictionary.cpp
@@ -26,6 +26,7 @@
 #include "suggest/core/dictionary/dictionary.h"
 #include "suggest/core/policy/dictionary_structure_with_buffer_policy.h"
 #include "suggest/core/result/suggestion_results.h"
+#include "suggest/core/session/prev_words_info.h"
 #include "utils/char_utils.h"
 
 namespace latinime {
@@ -42,19 +43,18 @@
 }
 
 /* Parameters :
- * prevWord: the word before, the one for which we need to look up bigrams.
- * prevWordLength: its length.
+ * prevWordsInfo: Information of previous words to get the predictions.
  * outSuggestionResults: SuggestionResults to put the predictions.
  */
-void BigramDictionary::getPredictions(const int *prevWord, const int prevWordLength,
+void BigramDictionary::getPredictions(const PrevWordsInfo *const prevWordsInfo,
         SuggestionResults *const outSuggestionResults) const {
-    int pos = getBigramListPositionForWord(prevWord, prevWordLength,
-            false /* forceLowerCaseSearch */);
+    int pos = getBigramListPositionForWord(prevWordsInfo->getPrevWordCodePoints(),
+            prevWordsInfo->getPrevWordCodePointCount(), false /* forceLowerCaseSearch */);
     // getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
     if (NOT_A_DICT_POS == pos) {
         // If no bigrams for this exact word, search again in lower case.
-        pos = getBigramListPositionForWord(prevWord, prevWordLength,
-                true /* forceLowerCaseSearch */);
+        pos = getBigramListPositionForWord(prevWordsInfo->getPrevWordCodePoints(),
+                prevWordsInfo->getPrevWordCodePointCount(), true /* forceLowerCaseSearch */);
     }
     // If still no bigrams, we really don't have them!
     if (NOT_A_DICT_POS == pos) return;
@@ -96,9 +96,10 @@
     return mDictionaryStructurePolicy->getBigramsPositionOfPtNode(pos);
 }
 
-int BigramDictionary::getBigramProbability(const int *word0, int length0, const int *word1,
-        int length1) const {
-    int pos = getBigramListPositionForWord(word0, length0, false /* forceLowerCaseSearch */);
+int BigramDictionary::getBigramProbability(const PrevWordsInfo *const prevWordsInfo,
+        const int *word1, int length1) const {
+    int pos = getBigramListPositionForWord(prevWordsInfo->getPrevWordCodePoints(),
+            prevWordsInfo->getPrevWordCodePointCount(), false /* forceLowerCaseSearch */);
     // getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
     if (NOT_A_DICT_POS == pos) return NOT_A_PROBABILITY;
     int nextWordPos = mDictionaryStructurePolicy->getTerminalPtNodePositionOfWord(word1, length1,
diff --git a/native/jni/src/suggest/core/dictionary/bigram_dictionary.h b/native/jni/src/suggest/core/dictionary/bigram_dictionary.h
index 12aaf20..bd3aed1 100644
--- a/native/jni/src/suggest/core/dictionary/bigram_dictionary.h
+++ b/native/jni/src/suggest/core/dictionary/bigram_dictionary.h
@@ -22,15 +22,17 @@
 namespace latinime {
 
 class DictionaryStructureWithBufferPolicy;
+class PrevWordsInfo;
 class SuggestionResults;
 
 class BigramDictionary {
  public:
     BigramDictionary(const DictionaryStructureWithBufferPolicy *const dictionaryStructurePolicy);
 
-    void getPredictions(const int *word, int length,
+    void getPredictions(const PrevWordsInfo *const prevWordsInfo,
             SuggestionResults *const outSuggestionResults) const;
-    int getBigramProbability(const int *word1, int length1, const int *word2, int length2) const;
+    int getBigramProbability(const PrevWordsInfo *const prevWordsInfo,
+            const int *word1, int length1) const;
     ~BigramDictionary();
 
  private:
diff --git a/native/jni/src/suggest/core/dictionary/dictionary.cpp b/native/jni/src/suggest/core/dictionary/dictionary.cpp
index aa52e63..c860d82 100644
--- a/native/jni/src/suggest/core/dictionary/dictionary.cpp
+++ b/native/jni/src/suggest/core/dictionary/dictionary.cpp
@@ -44,11 +44,11 @@
 
 void Dictionary::getSuggestions(ProximityInfo *proximityInfo, DicTraverseSession *traverseSession,
         int *xcoordinates, int *ycoordinates, int *times, int *pointerIds, int *inputCodePoints,
-        int inputSize, int *prevWordCodePoints, int prevWordLength,
+        int inputSize, const PrevWordsInfo *const prevWordsInfo,
         const SuggestOptions *const suggestOptions, const float languageWeight,
         SuggestionResults *const outSuggestionResults) const {
     TimeKeeper::setCurrentTime();
-    traverseSession->init(this, prevWordCodePoints, prevWordLength, suggestOptions);
+    traverseSession->init(this, prevWordsInfo, suggestOptions);
     const auto &suggest = suggestOptions->isGesture() ? mGestureSuggest : mTypingSuggest;
     suggest->getSuggestions(proximityInfo, traverseSession, xcoordinates,
             ycoordinates, times, pointerIds, inputCodePoints, inputSize,
@@ -58,11 +58,10 @@
     }
 }
 
-void Dictionary::getPredictions(const int *word, int length,
+void Dictionary::getPredictions(const PrevWordsInfo *const prevWordsInfo,
         SuggestionResults *const outSuggestionResults) const {
     TimeKeeper::setCurrentTime();
-    if (length <= 0) return;
-    mBigramDictionary.getPredictions(word, length, outSuggestionResults);
+    mBigramDictionary.getPredictions(prevWordsInfo, outSuggestionResults);
 }
 
 int Dictionary::getProbability(const int *word, int length) const {
@@ -75,10 +74,10 @@
     return getDictionaryStructurePolicy()->getUnigramProbabilityOfPtNode(pos);
 }
 
-int Dictionary::getBigramProbability(const int *word0, int length0, const int *word1,
+int Dictionary::getBigramProbability(const PrevWordsInfo *const prevWordsInfo, const int *word1,
         int length1) const {
     TimeKeeper::setCurrentTime();
-    return mBigramDictionary.getBigramProbability(word0, length0, word1, length1);
+    return mBigramDictionary.getBigramProbability(prevWordsInfo, word1, length1);
 }
 
 void Dictionary::addUnigramWord(const int *const word, const int length,
diff --git a/native/jni/src/suggest/core/dictionary/dictionary.h b/native/jni/src/suggest/core/dictionary/dictionary.h
index f0a7e5b..b63c61f 100644
--- a/native/jni/src/suggest/core/dictionary/dictionary.h
+++ b/native/jni/src/suggest/core/dictionary/dictionary.h
@@ -31,6 +31,7 @@
 
 class DictionaryStructureWithBufferPolicy;
 class DicTraverseSession;
+class PrevWordsInfo;
 class ProximityInfo;
 class SuggestionResults;
 class SuggestOptions;
@@ -62,16 +63,17 @@
 
     void getSuggestions(ProximityInfo *proximityInfo, DicTraverseSession *traverseSession,
             int *xcoordinates, int *ycoordinates, int *times, int *pointerIds, int *inputCodePoints,
-            int inputSize, int *prevWordCodePoints, int prevWordLength,
+            int inputSize, const PrevWordsInfo *const prevWordsInfo,
             const SuggestOptions *const suggestOptions, const float languageWeight,
             SuggestionResults *const outSuggestionResults) const;
 
-    void getPredictions(const int *word, int length,
+    void getPredictions(const PrevWordsInfo *const prevWordsInfo,
             SuggestionResults *const outSuggestionResults) const;
 
     int getProbability(const int *word, int length) const;
 
-    int getBigramProbability(const int *word0, int length0, const int *word1, int length1) const;
+    int getBigramProbability(const PrevWordsInfo *const prevWordsInfo,
+            const int *word1, int length1) const;
 
     void addUnigramWord(const int *const codePoints, const int codePointCount,
             const UnigramProperty *const unigramProperty);
diff --git a/native/jni/src/suggest/core/session/dic_traverse_session.cpp b/native/jni/src/suggest/core/session/dic_traverse_session.cpp
index 77b634e..7406fbe 100644
--- a/native/jni/src/suggest/core/session/dic_traverse_session.cpp
+++ b/native/jni/src/suggest/core/session/dic_traverse_session.cpp
@@ -20,6 +20,7 @@
 #include "suggest/core/dictionary/dictionary.h"
 #include "suggest/core/policy/dictionary_header_structure_policy.h"
 #include "suggest/core/policy/dictionary_structure_with_buffer_policy.h"
+#include "suggest/core/session/prev_words_info.h"
 
 namespace latinime {
 
@@ -28,24 +29,26 @@
 const int DicTraverseSession::DICTIONARY_SIZE_THRESHOLD_TO_USE_LARGE_CACHE_FOR_SUGGESTION =
         256 * 1024;
 
-void DicTraverseSession::init(const Dictionary *const dictionary, const int *prevWord,
-        int prevWordLength, const SuggestOptions *const suggestOptions) {
+void DicTraverseSession::init(const Dictionary *const dictionary,
+        const PrevWordsInfo *const prevWordsInfo, const SuggestOptions *const suggestOptions) {
     mDictionary = dictionary;
     mMultiWordCostMultiplier = getDictionaryStructurePolicy()->getHeaderStructurePolicy()
             ->getMultiWordCostMultiplier();
     mSuggestOptions = suggestOptions;
-    if (!prevWord) {
+    if (!prevWordsInfo->getPrevWordCodePoints()) {
         mPrevWordPtNodePos = NOT_A_DICT_POS;
         return;
     }
     // TODO: merge following similar calls to getTerminalPosition into one case-insensitive call.
     mPrevWordPtNodePos = getDictionaryStructurePolicy()->getTerminalPtNodePositionOfWord(
-            prevWord, prevWordLength, false /* forceLowerCaseSearch */);
+            prevWordsInfo->getPrevWordCodePoints(), prevWordsInfo->getPrevWordCodePointCount(),
+            false /* forceLowerCaseSearch */);
     if (mPrevWordPtNodePos == NOT_A_DICT_POS) {
         // Check bigrams for lower-cased previous word if original was not found. Useful for
         // auto-capitalized words like "The [current_word]".
         mPrevWordPtNodePos = getDictionaryStructurePolicy()->getTerminalPtNodePositionOfWord(
-                prevWord, prevWordLength, true /* forceLowerCaseSearch */);
+                prevWordsInfo->getPrevWordCodePoints(), prevWordsInfo->getPrevWordCodePointCount(),
+                true /* forceLowerCaseSearch */);
     }
 }
 
diff --git a/native/jni/src/suggest/core/session/dic_traverse_session.h b/native/jni/src/suggest/core/session/dic_traverse_session.h
index 50bffeb..50c60d0 100644
--- a/native/jni/src/suggest/core/session/dic_traverse_session.h
+++ b/native/jni/src/suggest/core/session/dic_traverse_session.h
@@ -29,6 +29,7 @@
 
 class Dictionary;
 class DictionaryStructureWithBufferPolicy;
+class PrevWordsInfo;
 class ProximityInfo;
 class SuggestOptions;
 
@@ -60,7 +61,7 @@
     // Non virtual inline destructor -- never inherit this class
     AK_FORCE_INLINE ~DicTraverseSession() {}
 
-    void init(const Dictionary *dictionary, const int *prevWord, int prevWordLength,
+    void init(const Dictionary *dictionary, const PrevWordsInfo *const prevWordsInfo,
             const SuggestOptions *const suggestOptions);
     // TODO: Remove and merge into init
     void setupForGetSuggestions(const ProximityInfo *pInfo, const int *inputCodePoints,
diff --git a/native/jni/src/suggest/core/session/prev_words_info.h b/native/jni/src/suggest/core/session/prev_words_info.h
new file mode 100644
index 0000000..df7e7e3
--- /dev/null
+++ b/native/jni/src/suggest/core/session/prev_words_info.h
@@ -0,0 +1,53 @@
+/*
+ * 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.
+ */
+
+#ifndef LATINIME_PREV_WORDS_INFO_H
+#define LATINIME_PREV_WORDS_INFO_H
+
+#include "defines.h"
+
+namespace latinime {
+
+// TODO: Support n-gram.
+// TODO: Support beginning of sentence.
+// This class does not take ownership of any code point buffers.
+class PrevWordsInfo {
+ public:
+    // No prev word information.
+    PrevWordsInfo()
+            : mPrevWordCodePoints(nullptr), mPrevWordCodePointCount(0) {}
+
+    PrevWordsInfo(const int *const prevWordCodePoints, const int prevWordCodePointCount,
+            const bool isBeginningOfSentence)
+            : mPrevWordCodePoints(prevWordCodePoints),
+              mPrevWordCodePointCount(prevWordCodePointCount) {}
+
+    const int *getPrevWordCodePoints() const {
+        return mPrevWordCodePoints;
+    }
+
+    int getPrevWordCodePointCount() const {
+        return mPrevWordCodePointCount;
+    }
+
+ private:
+    DISALLOW_COPY_AND_ASSIGN(PrevWordsInfo);
+
+    const int *const mPrevWordCodePoints;
+    const int mPrevWordCodePointCount;
+};
+} // namespace latinime
+#endif // LATINIME_PREV_WORDS_INFO_H