Support heteronyms in Chinese TTS (#738)

sherpa-onnx/csrc/CMakeLists.txt

@@ -164,6 +164,7 @@ endif()
 
 if(SHERPA_ONNX_ENABLE_TTS)
   target_link_libraries(sherpa-onnx-core piper_phonemize)
+  target_link_libraries(sherpa-onnx-core fstfar fst)
 endif()
 
 if(SHERPA_ONNX_ENABLE_CHECK)

sherpa-onnx/csrc/lexicon.cc

@@ -18,7 +18,6 @@
 #endif
 
 #include <memory>
-#include <regex>  // NOLINT
 
 #include "sherpa-onnx/csrc/macros.h"
 #include "sherpa-onnx/csrc/onnx-utils.h"
@@ -26,6 +25,55 @@
 
 namespace sherpa_onnx {
 
+static std::vector<std::string> ProcessHeteronyms(
+    const std::vector<std::string> &words) {
+  std::vector<std::string> ans;
+  ans.reserve(words.size());
+
+  int32_t num_words = static_cast<int32_t>(words.size());
+  int32_t i = 0;
+  int32_t prev = -1;
+  while (i < num_words) {
+    // start of a phrase #$|
+    if ((i + 2 < num_words) && words[i] == "#" && words[i + 1] == "$" &&
+        words[i + 2] == "|") {
+      if (prev == -1) {
+        prev = i + 3;
+      }
+      i = i + 3;
+      continue;
+    }
+
+    // end of a phrase |$#
+    if ((i + 2 < num_words) && words[i] == "|" && words[i + 1] == "$" &&
+        words[i + 2] == "#") {
+      if (prev != -1) {
+        std::ostringstream os;
+        for (int32_t k = prev; k < i; ++k) {
+          if (words[k] != "|" && words[k] != "$" && words[k] != "#") {
+            os << words[k];
+          }
+        }
+        ans.push_back(os.str());
+
+        prev = -1;
+      }
+
+      i += 3;
+      continue;
+    }
+
+    if (prev == -1) {
+      // not inside a phrase
+      ans.push_back(words[i]);
+    }
+
+    ++i;
+  }
+
+  return ans;
+}
+
 static void ToLowerCase(std::string *in_out) {
   std::transform(in_out->begin(), in_out->end(), in_out->begin(),
                  [](unsigned char c) { return std::tolower(c); });
@@ -148,36 +196,9 @@ std::vector<std::vector<int64_t>> Lexicon::ConvertTextToTokenIdsChinese(
     const std::string &_text) const {
   std::string text(_text);
   ToLowerCase(&text);
-  std::vector<std::string> words;
-  if (pattern_) {
-    // Handle polyphones
-    size_t pos = 0;
-    auto begin = std::sregex_iterator(text.begin(), text.end(), *pattern_);
-    auto end = std::sregex_iterator();
-    for (std::sregex_iterator i = begin; i != end; ++i) {
-      std::smatch match = *i;
-      if (pos < match.position()) {
-        auto this_segment = text.substr(pos, match.position() - pos);
-        auto this_segment_words = SplitUtf8(this_segment);
-        words.insert(words.end(), this_segment_words.begin(),
-                     this_segment_words.end());
-        pos = match.position() + match.length();
-      } else if (pos == match.position()) {
-        pos = match.position() + match.length();
-      }
 
-      words.push_back(match.str());
-    }
-
-    if (pos < text.size()) {
-      auto this_segment = text.substr(pos, text.size() - pos);
-      auto this_segment_words = SplitUtf8(this_segment);
-      words.insert(words.end(), this_segment_words.begin(),
-                   this_segment_words.end());
-    }
-  } else {
-    words = SplitUtf8(text);
-  }
+  std::vector<std::string> words = SplitUtf8(text);
+  words = ProcessHeteronyms(words);
 
   if (debug_) {
     fprintf(stderr, "Input text in string: %s\n", text.c_str());
@@ -357,9 +378,6 @@ void Lexicon::InitLexicon(std::istream &is) {
   std::string line;
   std::string phone;
 
-  std::ostringstream os;
-  std::string sep;
-
   while (std::getline(is, line)) {
     std::istringstream iss(line);
 
@@ -381,18 +399,9 @@ void Lexicon::InitLexicon(std::istream &is) {
     if (ids.empty()) {
       continue;
     }
-    if (language_ == Language::kChinese && word.size() > 3) {
-      // this is not a single word;
-      os << sep << word;
-      sep = "|";
-    }
 
     word2ids_.insert({std::move(word), std::move(ids)});
   }
-
-  if (!sep.empty()) {
-    pattern_ = std::make_unique<std::regex>(os.str());
-  }
 }
 
 void Lexicon::InitPunctuations(const std::string &punctuations) {

sherpa-onnx/csrc/lexicon.h

@@ -7,7 +7,6 @@
 
 #include <cstdint>
 #include <memory>
-#include <regex>  // NOLINT
 #include <string>
 #include <unordered_map>
 #include <unordered_set>
@@ -65,9 +64,6 @@ class Lexicon : public OfflineTtsFrontend {
   std::unordered_map<std::string, int32_t> token2id_;
   Language language_;
   bool debug_;
-
-  // for Chinese polyphones
-  std::unique_ptr<std::regex> pattern_;
 };
 
 }  // namespace sherpa_onnx

sherpa-onnx/csrc/offline-tts-vits-impl.h

@@ -15,6 +15,9 @@
 #include "android/asset_manager.h"
 #include "android/asset_manager_jni.h"
 #endif
+
+#include "fst/extensions/far/far.h"
+#include "kaldifst/csrc/kaldi-fst-io.h"
 #include "kaldifst/csrc/text-normalizer.h"
 #include "sherpa-onnx/csrc/lexicon.h"
 #include "sherpa-onnx/csrc/macros.h"
@@ -46,6 +49,32 @@ class OfflineTtsVitsImpl : public OfflineTtsImpl {
         tn_list_.push_back(std::make_unique<kaldifst::TextNormalizer>(f));
       }
     }
+
+    if (!config.rule_fars.empty()) {
+      if (config.model.debug) {
+        SHERPA_ONNX_LOGE("Loading FST archives");
+      }
+      std::vector<std::string> files;
+      SplitStringToVector(config.rule_fars, ",", false, &files);
+      for (const auto &f : files) {
+        if (config.model.debug) {
+          SHERPA_ONNX_LOGE("rule far: %s", f.c_str());
+        }
+        std::unique_ptr<fst::FarReader<fst::StdArc>> reader(
+            fst::FarReader<fst::StdArc>::Open(f));
+        for (; !reader->Done(); reader->Next()) {
+          std::unique_ptr<fst::StdConstFst> r(
+              fst::CastOrConvertToConstFst(reader->GetFst()->Copy()));
+
+          tn_list_.push_back(
+              std::make_unique<kaldifst::TextNormalizer>(std::move(r)));
+        }
+      }
+
+      if (config.model.debug) {
+        SHERPA_ONNX_LOGE("FST archives loaded!");
+      }
+    }
   }
 
 #if __ANDROID_API__ >= 9

sherpa-onnx/csrc/offline-tts.cc

@@ -20,7 +20,14 @@ void OfflineTtsConfig::Register(ParseOptions *po) {
                "It not empty, it contains a list of rule FST filenames."
                "Multiple filenames are separated by a comma and they are "
                "applied from left to right. An example value: "
-               "rule1.fst,rule2,fst,rule3.fst");
+               "rule1.fst,rule2.fst,rule3.fst");
+
+  po->Register("tts-rule-fars", &rule_fars,
+               "It not empty, it contains a list of rule FST archive filenames."
+               "Multiple filenames are separated by a comma and they are "
+               "applied from left to right. An example value: "
+               "rule1.far,rule2.far,rule3.far. Note that an *.far can contain "
+               "multiple *.fst files");
 
   po->Register(
       "tts-max-num-sentences", &max_num_sentences,
@@ -41,6 +48,17 @@ bool OfflineTtsConfig::Validate() const {
     }
   }
 
+  if (!rule_fars.empty()) {
+    std::vector<std::string> files;
+    SplitStringToVector(rule_fars, ",", false, &files);
+    for (const auto &f : files) {
+      if (!FileExists(f)) {
+        SHERPA_ONNX_LOGE("Rule far %s does not exist. ", f.c_str());
+        return false;
+      }
+    }
+  }
+
   return model.Validate();
 }
 
@@ -50,6 +68,7 @@ std::string OfflineTtsConfig::ToString() const {
   os << "OfflineTtsConfig(";
   os << "model=" << model.ToString() << ", ";
   os << "rule_fsts=\"" << rule_fsts << "\", ";
+  os << "rule_fars=\"" << rule_fars << "\", ";
   os << "max_num_sentences=" << max_num_sentences << ")";
 
   return os.str();

sherpa-onnx/csrc/offline-tts.h

@@ -29,6 +29,9 @@ struct OfflineTtsConfig {
   // If there are multiple rules, they are applied from left to right.
   std::string rule_fsts;
 
+  // If there are multiple FST archives, they are applied from left to right.
+  std::string rule_fars;
+
   // Maximum number of sentences that we process at a time.
   // This is to avoid OOM for very long input text.
   // If you set it to -1, then we process all sentences in a single batch.
@@ -36,9 +39,11 @@ struct OfflineTtsConfig {
 
   OfflineTtsConfig() = default;
   OfflineTtsConfig(const OfflineTtsModelConfig &model,
-                   const std::string &rule_fsts, int32_t max_num_sentences)
+                   const std::string &rule_fsts, const std::string &rule_fars,
+                   int32_t max_num_sentences)
       : model(model),
         rule_fsts(rule_fsts),
+        rule_fars(rule_fars),
         max_num_sentences(max_num_sentences) {}
 
   void Register(ParseOptions *po);