Lines Matching refs:bigram
96 * bigram address list is:
496 // extreme overhead at bigram lookup time (it would make the search function O(n) instead
556 for (WeightedString bigram : group.mBigrams) {
559 final int addressOfBigram = findAddressOfWord(dict, bigram.mWord);
727 throw new RuntimeException("0-sized bigram list must be null");
735 * Makes the flag value for a bigram.
738 * @param offset the offset of the bigram.
739 * @param bigramFrequency the frequency of the bigram, 0..255.
741 * @param word the second bigram, for debugging purposes
762 MakedictLog.e("Unigram freq is superior to bigram freq for \"" + word
763 + "\". Bigram freq is " + bigramFrequency + ", unigram freq for "
771 // value of the 16th step. In all justice, if the bigram frequency is low enough to be
774 // bigram freq that we could possibly supply, hence we should *not* include this bigram
794 // If the bigram freq is less than half-a-step higher than the unigram freq, we get -1
796 // include this bigram in the dictionary. For now, register as 0, and live with the
797 // small over-estimation that we get in this case. TODO: actually remove this bigram
907 final WeightedString bigram = bigramIterator.next();
909 FusionDictionary.findWordInTree(dict.mRoot, bigram.mWord);
915 bigram.mFrequency, unigramFrequencyForThisWord, bigram.mWord);
1291 for (PendingAttribute bigram : info.mBigrams) {
1292 final String word = getWordAtAddress(source, headerSize, bigram.mAddress);
1293 bigrams.add(new WeightedString(word, bigram.mFrequency));
1385 for (final WeightedString bigram : w.mBigrams) {
1386 newDict.setBigram(w.mWord, bigram.mWord, bigram.mFrequency);
