Add blank penalty for various language bindings. (#1234)

k2-fsa · Aug 8, 2024 · 94e2562 · 94e2562
1 parent ba4cb61
commit 94e2562
Show file tree

Hide file tree

Showing 38 changed files with 123 additions and 42 deletions.
diff --git a/.gitignore b/.gitignore
@@ -113,3 +113,4 @@ sherpa-onnx-telespeech-ctc-*
 lib*.a
 sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17
 *.bak
+vits-melo-tts-zh_en
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,3 +1,22 @@
+## 1.10.21
+
+* Fix ffmpeg c api example (#1185)
+* Fix splitting sentences for MeloTTS (#1186)
+* Non-streaming WebSocket client for Java. (#1190)
+* Fix copying asset files for flutter examples. (#1191)
+* Add Chinese+English tts example for flutter (#1192)
+* Add speaker identification and verification exmaple for Dart API (#1194)
+* Fix reading non-standard wav files. (#1199)
+* Add ReazonSpeech Japanese pre-trained model (#1203)
+* Describe how to add new words for MeloTTS models (#1209)
+* Remove libonnxruntime_providers_cuda.so as a dependency. (#1210)
+* Fix setting SenseVoice language. (#1214)
+* Support passing TTS callback in Swift API (#1218)
+* Add MeloTTS example for ios (#1223)
+* Add online punctuation and casing prediction model for English language (#1224)
+* Fix python two pass ASR examples (#1230)
+* Add blank penalty for various language bindings
+
 ## 1.10.20
 
 * Add Dart API for audio tagging

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -11,7 +11,7 @@ project(sherpa-onnx)
 # ./nodejs-addon-examples
 # ./dart-api-examples/
 # ./CHANGELOG.md
-set(SHERPA_ONNX_VERSION "1.10.20")
+set(SHERPA_ONNX_VERSION "1.10.21")
 
 # Disable warning about
 #

diff --git a/build-ios-no-tts.sh b/build-ios-no-tts.sh
@@ -126,7 +126,7 @@ echo "Generate xcframework"
 
 mkdir -p "build/simulator/lib"
 for f in libkaldi-native-fbank-core.a libsherpa-onnx-c-api.a libsherpa-onnx-core.a \
-         libsherpa-onnx-fst.a libsherpa-onnx-kaldifst-core.a libkaldi-decoder-core.a libssentencepiece_core.a; do
+         libsherpa-onnx-fst.a libsherpa-onnx-fstfar.a libsherpa-onnx-kaldifst-core.a libkaldi-decoder-core.a libssentencepiece_core.a; do
   lipo -create build/simulator_arm64/lib/${f} \
                build/simulator_x86_64/lib/${f} \
        -output build/simulator/lib/${f}
@@ -139,6 +139,7 @@ libtool -static -o build/simulator/sherpa-onnx.a \
   build/simulator/lib/libsherpa-onnx-c-api.a \
   build/simulator/lib/libsherpa-onnx-core.a  \
   build/simulator/lib/libsherpa-onnx-fst.a   \
+  build/simulator/lib/libsherpa-onnx-fstfar.a   \
   build/simulator/lib/libsherpa-onnx-kaldifst-core.a \
   build/simulator/lib/libkaldi-decoder-core.a \
   build/simulator/lib/libssentencepiece_core.a
@@ -148,6 +149,7 @@ libtool -static -o build/os64/sherpa-onnx.a \
   build/os64/lib/libsherpa-onnx-c-api.a \
   build/os64/lib/libsherpa-onnx-core.a \
   build/os64/lib/libsherpa-onnx-fst.a   \
+  build/os64/lib/libsherpa-onnx-fstfar.a   \
   build/os64/lib/libsherpa-onnx-kaldifst-core.a \
   build/os64/lib/libkaldi-decoder-core.a \
   build/os64/lib/libssentencepiece_core.a

diff --git a/dart-api-examples/add-punctuations/pubspec.yaml b/dart-api-examples/add-punctuations/pubspec.yaml
@@ -9,7 +9,7 @@ environment:
   sdk: ^3.4.0
 
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/dart-api-examples/audio-tagging/pubspec.yaml b/dart-api-examples/audio-tagging/pubspec.yaml
@@ -9,7 +9,7 @@ environment:
   sdk: ^3.4.0
 
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/dart-api-examples/keyword-spotter/pubspec.yaml b/dart-api-examples/keyword-spotter/pubspec.yaml
@@ -9,7 +9,7 @@ environment:
   sdk: ^3.4.0
 
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   # sherpa_onnx:
   #   path: ../../flutter/sherpa_onnx
   path: ^1.9.0

diff --git a/dart-api-examples/non-streaming-asr/pubspec.yaml b/dart-api-examples/non-streaming-asr/pubspec.yaml
@@ -10,7 +10,7 @@ environment:
 
 # Add regular dependencies here.
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/dart-api-examples/speaker-identification/pubspec.yaml b/dart-api-examples/speaker-identification/pubspec.yaml
@@ -9,7 +9,7 @@ environment:
   sdk: ^3.4.0
 
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/dart-api-examples/streaming-asr/pubspec.yaml b/dart-api-examples/streaming-asr/pubspec.yaml
@@ -11,7 +11,7 @@ environment:
 
 # Add regular dependencies here.
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/dart-api-examples/tts/pubspec.yaml b/dart-api-examples/tts/pubspec.yaml
@@ -8,7 +8,7 @@ environment:
 
 # Add regular dependencies here.
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/dart-api-examples/vad-with-non-streaming-asr/pubspec.yaml b/dart-api-examples/vad-with-non-streaming-asr/pubspec.yaml
@@ -10,7 +10,7 @@ environment:
   sdk: ^3.4.0
 
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/dart-api-examples/vad/pubspec.yaml b/dart-api-examples/vad/pubspec.yaml
@@ -9,7 +9,7 @@ environment:
   sdk: ^3.4.0
 
 dependencies:
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   path: ^1.9.0
   args: ^2.5.0
 

diff --git a/flutter-examples/streaming_asr/pubspec.yaml b/flutter-examples/streaming_asr/pubspec.yaml
@@ -5,7 +5,7 @@ description: >
 
 publish_to: 'none'
 
-version: 1.10.20
+version: 1.10.21
 
 topics:
   - speech-recognition
@@ -30,7 +30,7 @@ dependencies:
   record: ^5.1.0
   url_launcher: ^6.2.6
 
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   # sherpa_onnx:
     # path: ../../flutter/sherpa_onnx
 

diff --git a/flutter-examples/tts/pubspec.yaml b/flutter-examples/tts/pubspec.yaml
@@ -5,7 +5,7 @@ description: >
 
 publish_to: 'none' # Remove this line if you wish to publish to pub.dev
 
-version: 1.10.20
+version: 1.10.21
 
 environment:
   sdk: '>=3.4.0 <4.0.0'
@@ -17,7 +17,7 @@ dependencies:
   cupertino_icons: ^1.0.6
   path_provider: ^2.1.3
   path: ^1.9.0
-  sherpa_onnx: ^1.10.20
+  sherpa_onnx: ^1.10.21
   url_launcher: ^6.2.6
   audioplayers: ^5.0.0
 

diff --git a/flutter/sherpa_onnx/lib/src/offline_recognizer.dart b/flutter/sherpa_onnx/lib/src/offline_recognizer.dart
@@ -159,11 +159,12 @@ class OfflineRecognizerConfig {
     this.hotwordsScore = 1.5,
     this.ruleFsts = '',
     this.ruleFars = '',
+    this.blankPenalty = 0.0,
   });
 
   @override
   String toString() {
-    return 'OfflineRecognizerConfig(feat: $feat, model: $model, lm: $lm, decodingMethod: $decodingMethod, maxActivePaths: $maxActivePaths, hotwordsFile: $hotwordsFile, hotwordsScore: $hotwordsScore, ruleFsts: $ruleFsts, ruleFars: $ruleFars)';
+    return 'OfflineRecognizerConfig(feat: $feat, model: $model, lm: $lm, decodingMethod: $decodingMethod, maxActivePaths: $maxActivePaths, hotwordsFile: $hotwordsFile, hotwordsScore: $hotwordsScore, ruleFsts: $ruleFsts, ruleFars: $ruleFars, blankPenalty: $blankPenalty)';
   }
 
   final FeatureConfig feat;
@@ -179,6 +180,8 @@ class OfflineRecognizerConfig {
 
   final String ruleFsts;
   final String ruleFars;
+
+  final double blankPenalty;
 }
 
 class OfflineRecognizerResult {
@@ -268,6 +271,8 @@ class OfflineRecognizer {
     c.ref.ruleFsts = config.ruleFsts.toNativeUtf8();
     c.ref.ruleFars = config.ruleFars.toNativeUtf8();
 
+    c.ref.blankPenalty = config.blankPenalty;
+
     final ptr = SherpaOnnxBindings.createOfflineRecognizer?.call(c) ?? nullptr;
 
     calloc.free(c.ref.ruleFars);

diff --git a/flutter/sherpa_onnx/lib/src/online_recognizer.dart b/flutter/sherpa_onnx/lib/src/online_recognizer.dart
@@ -114,11 +114,12 @@ class OnlineRecognizerConfig {
     this.ctcFstDecoderConfig = const OnlineCtcFstDecoderConfig(),
     this.ruleFsts = '',
     this.ruleFars = '',
+    this.blankPenalty = 0.0,
   });
 
   @override
   String toString() {
-    return 'OnlineRecognizerConfig(feat: $feat, model: $model, decodingMethod: $decodingMethod, maxActivePaths: $maxActivePaths, enableEndpoint: $enableEndpoint, rule1MinTrailingSilence: $rule1MinTrailingSilence, rule2MinTrailingSilence: $rule2MinTrailingSilence, rule3MinUtteranceLength: $rule3MinUtteranceLength, hotwordsFile: $hotwordsFile, hotwordsScore: $hotwordsScore, ctcFstDecoderConfig: $ctcFstDecoderConfig, ruleFsts: $ruleFsts, ruleFars: $ruleFars)';
+    return 'OnlineRecognizerConfig(feat: $feat, model: $model, decodingMethod: $decodingMethod, maxActivePaths: $maxActivePaths, enableEndpoint: $enableEndpoint, rule1MinTrailingSilence: $rule1MinTrailingSilence, rule2MinTrailingSilence: $rule2MinTrailingSilence, rule3MinUtteranceLength: $rule3MinUtteranceLength, hotwordsFile: $hotwordsFile, hotwordsScore: $hotwordsScore, ctcFstDecoderConfig: $ctcFstDecoderConfig, ruleFsts: $ruleFsts, ruleFars: $ruleFars, blankPenalty: $blankPenalty)';
   }
 
   final FeatureConfig feat;
@@ -142,6 +143,8 @@ class OnlineRecognizerConfig {
   final OnlineCtcFstDecoderConfig ctcFstDecoderConfig;
   final String ruleFsts;
   final String ruleFars;
+
+  final double blankPenalty;
 }
 
 class OnlineRecognizerResult {
@@ -209,6 +212,8 @@ class OnlineRecognizer {
     c.ref.ruleFsts = config.ruleFsts.toNativeUtf8();
     c.ref.ruleFars = config.ruleFars.toNativeUtf8();
 
+    c.ref.blankPenalty = config.blankPenalty;
+
     final ptr = SherpaOnnxBindings.createOnlineRecognizer?.call(c) ?? nullptr;
 
     calloc.free(c.ref.ruleFars);

diff --git a/flutter/sherpa_onnx/lib/src/sherpa_onnx_bindings.dart b/flutter/sherpa_onnx/lib/src/sherpa_onnx_bindings.dart
@@ -194,6 +194,9 @@ final class SherpaOnnxOfflineRecognizerConfig extends Struct {
 
   external Pointer<Utf8> ruleFsts;
   external Pointer<Utf8> ruleFars;
+
+  @Float()
+  external double blankPenalty;
 }
 
 final class SherpaOnnxOnlineTransducerModelConfig extends Struct {
@@ -269,6 +272,9 @@ final class SherpaOnnxOnlineRecognizerConfig extends Struct {
 
   external Pointer<Utf8> ruleFsts;
   external Pointer<Utf8> ruleFars;
+
+  @Float()
+  external double blankPenalty;
 }
 
 final class SherpaOnnxSileroVadModelConfig extends Struct {

diff --git a/flutter/sherpa_onnx/pubspec.yaml b/flutter/sherpa_onnx/pubspec.yaml
@@ -17,7 +17,7 @@ topics:
   - voice-activity-detection
 
 # remember to change the version in ../sherpa_onnx_macos/macos/sherpa_onnx_macos.podspec
-version: 1.10.20
+version: 1.10.21
 
 homepage: https://github.com/k2-fsa/sherpa-onnx
 
@@ -30,23 +30,23 @@ dependencies:
   flutter:
     sdk: flutter
 
-  sherpa_onnx_android: ^1.10.20
+  sherpa_onnx_android: ^1.10.21
   # sherpa_onnx_android:
   #   path: ../sherpa_onnx_android
 
-  sherpa_onnx_macos: ^1.10.20
+  sherpa_onnx_macos: ^1.10.21
   # sherpa_onnx_macos:
   #   path: ../sherpa_onnx_macos
 
-  sherpa_onnx_linux: ^1.10.20
+  sherpa_onnx_linux: ^1.10.21
   # sherpa_onnx_linux:
   #   path: ../sherpa_onnx_linux
     #
-  sherpa_onnx_windows: ^1.10.20
+  sherpa_onnx_windows: ^1.10.21
   # sherpa_onnx_windows:
   #   path: ../sherpa_onnx_windows
 
-  sherpa_onnx_ios: ^1.10.20
+  sherpa_onnx_ios: ^1.10.21
   # sherpa_onnx_ios:
   #   path: ../sherpa_onnx_ios
 

diff --git a/flutter/sherpa_onnx_ios/ios/sherpa_onnx_ios.podspec b/flutter/sherpa_onnx_ios/ios/sherpa_onnx_ios.podspec
@@ -7,7 +7,7 @@
 # https://groups.google.com/g/dart-ffi/c/nUATMBy7r0c
 Pod::Spec.new do |s|
   s.name             = 'sherpa_onnx_ios'
-  s.version          = '1.10.20'
+  s.version          = '1.10.21'
   s.summary          = 'A new Flutter FFI plugin project.'
   s.description      = <<-DESC
 A new Flutter FFI plugin project.

diff --git a/flutter/sherpa_onnx_macos/macos/sherpa_onnx_macos.podspec b/flutter/sherpa_onnx_macos/macos/sherpa_onnx_macos.podspec
@@ -4,7 +4,7 @@
 #
 Pod::Spec.new do |s|
   s.name             = 'sherpa_onnx_macos'
-  s.version          = '1.10.20'
+  s.version          = '1.10.21'
   s.summary          = 'sherpa-onnx Flutter FFI plugin project.'
   s.description      = <<-DESC
 sherpa-onnx Flutter FFI plugin project.

diff --git a/nodejs-addon-examples/package.json b/nodejs-addon-examples/package.json
@@ -1,5 +1,5 @@
 {
   "dependencies": {
-    "sherpa-onnx-node": "^1.10.20"
+    "sherpa-onnx-node": "^1.10.21"
   }
 }
diff --git a/scripts/dotnet/OfflineRecognizerConfig.cs b/scripts/dotnet/OfflineRecognizerConfig.cs
@@ -4,7 +4,6 @@
 
 namespace SherpaOnnx
 {
-
     [StructLayout(LayoutKind.Sequential)]
     public struct OfflineRecognizerConfig
     {
@@ -20,6 +19,7 @@ public OfflineRecognizerConfig()
             HotwordsScore = 1.5F;
             RuleFsts = "";
             RuleFars = "";
+            BlankPenalty = 0.0F;
         }
         public FeatureConfig FeatConfig;
         public OfflineModelConfig ModelConfig;
@@ -40,7 +40,7 @@ public OfflineRecognizerConfig()
 
         [MarshalAs(UnmanagedType.LPStr)]
         public string RuleFars;
-    }
 
-
-}
+        public float BlankPenalty;
+    }
+}
diff --git a/scripts/dotnet/OnlineRecognizerConfig.cs b/scripts/dotnet/OnlineRecognizerConfig.cs
@@ -25,6 +25,7 @@ public OnlineRecognizerConfig()
             CtcFstDecoderConfig = new OnlineCtcFstDecoderConfig();
             RuleFsts = "";
             RuleFars = "";
+            BlankPenalty = 0.0F;
         }
         public FeatureConfig FeatConfig;
         public OnlineModelConfig ModelConfig;
@@ -69,6 +70,7 @@ public OnlineRecognizerConfig()
 
         [MarshalAs(UnmanagedType.LPStr)]
         public string RuleFars;
-    }
 
-}
+        public float BlankPenalty;
+    }
+}
diff --git a/scripts/node-addon-api/src/non-streaming-asr.cc b/scripts/node-addon-api/src/non-streaming-asr.cc
@@ -201,6 +201,7 @@ CreateOfflineRecognizerWrapper(const Napi::CallbackInfo &info) {
   SHERPA_ONNX_ASSIGN_ATTR_FLOAT(hotwords_score, hotwordsScore);
   SHERPA_ONNX_ASSIGN_ATTR_STR(rule_fsts, ruleFsts);
   SHERPA_ONNX_ASSIGN_ATTR_STR(rule_fars, ruleFars);
+  SHERPA_ONNX_ASSIGN_ATTR_FLOAT(blank_penalty, blankPenalty);
 
   SherpaOnnxOfflineRecognizer *recognizer =
       SherpaOnnxCreateOfflineRecognizer(&c);

diff --git a/scripts/node-addon-api/src/streaming-asr.cc b/scripts/node-addon-api/src/streaming-asr.cc
@@ -191,6 +191,7 @@ static Napi::External<SherpaOnnxOnlineRecognizer> CreateOnlineRecognizerWrapper(
   SHERPA_ONNX_ASSIGN_ATTR_FLOAT(hotwords_score, hotwordsScore);
   SHERPA_ONNX_ASSIGN_ATTR_STR(rule_fsts, ruleFsts);
   SHERPA_ONNX_ASSIGN_ATTR_STR(rule_fars, ruleFars);
+  SHERPA_ONNX_ASSIGN_ATTR_FLOAT(blank_penalty, blankPenalty);
 
   c.ctc_fst_decoder_config = GetCtcFstDecoderConfig(o);