5 files changed, 33 insertions, 4 deletions

diff --git a/native/jni/src/bigram_dictionary.cpp b/native/jni/src/bigram_dictionary.cpp
index 07031086c..7ed4dc439 100644
--- a/native/jni/src/bigram_dictionary.cpp
+++ b/native/jni/src/bigram_dictionary.cpp
@@ -128,7 +128,7 @@ int BigramDictionary::getBigrams(const int32_t *prevWord, int prevWordLength, in
                 ++bigramCount;
             }
         }
-    } while (0 != (UnigramDictionary::FLAG_ATTRIBUTE_HAS_NEXT & bigramFlags));
+    } while (UnigramDictionary::FLAG_ATTRIBUTE_HAS_NEXT & bigramFlags);
     return bigramCount;
 }
 
@@ -189,5 +189,25 @@ bool BigramDictionary::checkFirstCharacter(unsigned short *word) {
     return false;
 }
 
+bool BigramDictionary::isValidBigram(const int32_t *word1, int length1, const int32_t *word2,
+        int length2) {
+    const uint8_t* const root = DICT;
+    int pos = getBigramListPositionForWord(word1, length1);
+    // getBigramListPositionForWord returns 0 if this word isn't in the dictionary or has no bigrams
+    if (0 == pos) return false;
+    int nextWordPos = BinaryFormat::getTerminalPosition(root, word2, length2);
+    if (NOT_VALID_WORD == nextWordPos) return false;
+    int bigramFlags;
+    do {
+        bigramFlags = BinaryFormat::getFlagsAndForwardPointer(root, &pos);
+        const int bigramPos = BinaryFormat::getAttributeAddressAndForwardPointer(root, bigramFlags,
+                &pos);
+        if (bigramPos == nextWordPos) {
+            return true;
+        }
+    } while (UnigramDictionary::FLAG_ATTRIBUTE_HAS_NEXT & bigramFlags);
+    return false;
+}
+
 // TODO: Move functions related to bigram to here
 } // namespace latinime
diff --git a/native/jni/src/bigram_dictionary.h b/native/jni/src/bigram_dictionary.h
index 7328d5828..b8763a515 100644
--- a/native/jni/src/bigram_dictionary.h
+++ b/native/jni/src/bigram_dictionary.h
@@ -33,6 +33,7 @@ class BigramDictionary {
     int getBigramListPositionForWord(const int32_t *prevWord, const int prevWordLength);
     void fillBigramAddressToFrequencyMapAndFilter(const int32_t *prevWord, const int prevWordLength,
             std::map<int, int> *map, uint8_t *filter);
+    bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2);
     ~BigramDictionary();
  private:
     bool addWordBigram(unsigned short *word, int length, int frequency);
diff --git a/native/jni/src/correction.cpp b/native/jni/src/correction.cpp
index 376e9a10e..a1f812909 100644
--- a/native/jni/src/correction.cpp
+++ b/native/jni/src/correction.cpp
@@ -344,8 +344,10 @@ Correction::CorrectionType Correction::processCharAndCalcState(
                 mDistances[mOutputIndex] =
                         mProximityInfo->getNormalizedSquaredDistance(mInputIndex, proximityIndex);
             }
-            incrementInputIndex();
-            incremented = true;
+            if (!isQuote(c)) {
+                incrementInputIndex();
+                incremented = true;
+            }
         }
         return processSkipChar(c, isTerminal, incremented);
     }
@@ -710,7 +712,7 @@ int Correction::RankingAlgorithm::calculateFinalProbability(const int inputIndex
         ed = max(0, ed - quoteDiffCount);
         adjustedProximityMatchedCount = min(max(0, ed - (outputLength - inputLength)),
                 proximityMatchedCount);
-        if (transposedCount < 1) {
+        if (transposedCount <= 0) {
             if (ed == 1 && (inputLength == outputLength - 1 || inputLength == outputLength + 1)) {
                 // Promote a word with just one skipped or excessive char
                 if (sameLength) {
diff --git a/native/jni/src/dictionary.cpp b/native/jni/src/dictionary.cpp
index 9dc207223..8ea7c49fa 100644
--- a/native/jni/src/dictionary.cpp
+++ b/native/jni/src/dictionary.cpp
@@ -58,4 +58,9 @@ bool Dictionary::isValidWord(const int32_t *word, int length) {
     return mUnigramDictionary->isValidWord(word, length);
 }
 
+bool Dictionary::isValidBigram(const int32_t *word1, int length1, const int32_t *word2,
+        int length2) {
+    return mBigramDictionary->isValidBigram(word1, length1, word2, length2);
+}
+
 } // namespace latinime
diff --git a/native/jni/src/dictionary.h b/native/jni/src/dictionary.h
index bce86d1ad..87891ee4d 100644
--- a/native/jni/src/dictionary.h
+++ b/native/jni/src/dictionary.h
@@ -53,6 +53,7 @@ class Dictionary {
     }
 
     bool isValidWord(const int32_t *word, int length);
+    bool isValidBigram(const int32_t *word1, int length1, const int32_t *word2, int length2);
     void *getDict() { return (void *)mDict; }
     int getDictSize() { return mDictSize; }
     int getMmapFd() { return mMmapFd; }