Merge pull request #2663 from bertsky/fix-lstm-user-patterns

fix langdata (user words/patterns) file suffixes for LSTMs:

Author: zdenop

src/ccmain/tessedit.cpp

@@ -175,7 +175,7 @@ bool Tesseract::init_tesseract_lang_data(
       tessedit_ocr_engine_mode == OEM_TESSERACT_LSTM_COMBINED) {
 #  endif  // ndef DISABLED_LEGACY_ENGINE
     if (mgr->IsComponentAvailable(TESSDATA_LSTM)) {
-      lstm_recognizer_ = new LSTMRecognizer;
+      lstm_recognizer_ = new LSTMRecognizer(language_data_path_prefix);
       ASSERT_HOST(lstm_recognizer_->Load(
           this->params(), lstm_use_matrix ? language : nullptr, mgr));
     } else {

src/lstm/lstmrecognizer.cpp

@@ -49,6 +49,11 @@ const double kDictRatio = 2.25;
 // Default certainty offset to give the dictionary a chance.
 const double kCertOffset = -0.085;
 
+LSTMRecognizer::LSTMRecognizer(const STRING language_data_path_prefix)
+    : LSTMRecognizer::LSTMRecognizer() {
+  ccutil_.language_data_path_prefix = language_data_path_prefix;
+}
+
 LSTMRecognizer::LSTMRecognizer()
     : network_(nullptr),
       training_flags_(0),

src/lstm/lstmrecognizer.h

@@ -53,6 +53,7 @@ enum TrainingFlags {
 class LSTMRecognizer {
  public:
   LSTMRecognizer();
+  LSTMRecognizer(const STRING language_data_path_prefix);
   ~LSTMRecognizer();
 
   int NumOutputs() const { return network_->NumOutputs(); }