Inverse text normalization API for other programming languages

.github/scripts/test-dart.sh

@@ -4,6 +4,41 @@ set -ex
 
 cd dart-api-examples
 
+pushd non-streaming-asr
+
+echo '----------paraformer itn----------'
+./run-paraformer-itn.sh
+
+echo '----------paraformer----------'
+./run-paraformer.sh
+rm -rf sherpa-onnx-*
+
+echo '----------VAD with paraformer----------'
+./run-vad-with-paraformer.sh
+rm -rf sherpa-onnx-*
+
+echo '----------NeMo transducer----------'
+./run-nemo-transducer.sh
+rm -rf sherpa-onnx-*
+
+echo '----------NeMo CTC----------'
+./run-nemo-ctc.sh
+rm -rf sherpa-onnx-*
+
+echo '----------TeleSpeech CTC----------'
+./run-telespeech-ctc.sh
+rm -rf sherpa-onnx-*
+
+echo '----------whisper----------'
+./run-whisper.sh
+rm -rf sherpa-onnx-*
+
+echo '----------zipformer transducer----------'
+./run-zipformer-transducer.sh
+rm -rf sherpa-onnx-*
+
+popd # non-streaming-asr
+
 pushd tts
 
 echo '----------piper tts----------'
@@ -44,38 +79,6 @@ rm -rf sherpa-onnx-*
 
 popd # streaming-asr
 
-pushd non-streaming-asr
-
-echo '----------VAD with paraformer----------'
-./run-vad-with-paraformer.sh
-rm -rf sherpa-onnx-*
-
-echo '----------NeMo transducer----------'
-./run-nemo-transducer.sh
-rm -rf sherpa-onnx-*
-
-echo '----------NeMo CTC----------'
-./run-nemo-ctc.sh
-rm -rf sherpa-onnx-*
-
-echo '----------TeleSpeech CTC----------'
-./run-telespeech-ctc.sh
-rm -rf sherpa-onnx-*
-
-echo '----------paraformer----------'
-./run-paraformer.sh
-rm -rf sherpa-onnx-*
-
-echo '----------whisper----------'
-./run-whisper.sh
-rm -rf sherpa-onnx-*
-
-echo '----------zipformer transducer----------'
-./run-zipformer-transducer.sh
-rm -rf sherpa-onnx-*
-
-popd # non-streaming-asr
-
 pushd vad
 ./run.sh
 rm *.onnx

.github/scripts/test-dot-net.sh

@@ -3,6 +3,7 @@
 cd dotnet-examples/
 
 cd ./offline-decode-files
+./run-paraformer-itn.sh
 ./run-telespeech-ctc.sh
 ./run-nemo-ctc.sh
 ./run-paraformer.sh

.github/scripts/test-nodejs-addon-npm.sh

@@ -119,6 +119,12 @@ tar xvf sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
 rm sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
 
 node ./test_asr_non_streaming_paraformer.js
+
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn_zh_number.fst
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn-zh-number.wav
+
+node ./test_asr_non_streaming_paraformer_itn.js
+
 rm -rf sherpa-onnx-paraformer-zh-2023-03-28
 
 echo "----------tts----------"

.github/scripts/test-nodejs-npm.sh

@@ -11,6 +11,15 @@ ls -lh node_modules
 
 # offline asr
 
+curl -LS -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+ls -lh
+tar xvf sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+rm sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn-zh-number.wav
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn_zh_number.fst
+node ./test-offline-paraformer-itn.js
+rm -rf sherpa-onnx-paraformer-zh-2023-03-28
+
 curl -LS -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-ctc-en-conformer-small.tar.bz2
 ls -lh
 tar xvf sherpa-onnx-nemo-ctc-en-conformer-small.tar.bz2

.github/workflows/run-java-test.yaml

@@ -190,6 +190,8 @@ jobs:
         shell: bash
         run: |
           cd ./java-api-examples
+          ./run-inverse-text-normalization-paraformer.sh
+
           ./run-non-streaming-decode-file-paraformer.sh
           rm -rf sherpa-onnx-paraformer-zh-*
 

.github/workflows/test-dot-net.yaml

@@ -39,7 +39,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
+        os: [ubuntu-latest]
         python-version: ["3.8"]
 
     steps:
@@ -72,45 +72,18 @@ jobs:
 
           cmake --build . --target install --config Release
 
-      - name: Build sherpa-onnx for windows x86
-        if: matrix.os == 'windows-latest'
-        shell: bash
-        run: |
-          export CMAKE_CXX_COMPILER_LAUNCHER=ccache
-          export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
-          cmake --version
-
-          mkdir build-win32
-          cd build-win32
-          cmake \
-            -A Win32 \
-            -DBUILD_SHARED_LIBS=ON \
-            -DCMAKE_INSTALL_PREFIX=./install \
-            -DCMAKE_BUILD_TYPE=Release \
-            -DSHERPA_ONNX_ENABLE_WEBSOCKET=OFF \
-            -DBUILD_ESPEAK_NG_EXE=OFF \
-            -DSHERPA_ONNX_ENABLE_BINARY=ON \
-            ..
-          cmake --build . --target install --config Release
-
       - uses: actions/upload-artifact@v4
         with:
           name: ${{ matrix.os }}
           path: ./build/install/lib/
 
-      - uses: actions/upload-artifact@v4
-        if: matrix.os == 'windows-latest'
-        with:
-          name: ${{ matrix.os }}-win32
-          path: ./build-win32/install/lib/
-
   test-dot-net:
     runs-on: ${{ matrix.os }}
     needs: [build-libs]
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-latest] #, windows-latest]
+        os: [ubuntu-latest]
         python-version: ["3.8"]
 
     steps:
@@ -134,30 +107,11 @@ jobs:
           name: ubuntu-latest
           path: /tmp/linux
 
-      - name: Retrieve artifact from macos-latest
-        uses: actions/download-artifact@v4
-        with:
-          name: macos-latest
-          path: /tmp/macos
-
-      - name: Retrieve artifact from windows-latest
-        uses: actions/download-artifact@v4
-        with:
-          name: windows-latest
-          path: /tmp/windows-x64
-
-      - name: Retrieve artifact from windows-latest
-        uses: actions/download-artifact@v4
-        with:
-          name: windows-latest-win32
-          path: /tmp/windows-x86
-
       - name: Setup .NET
         uses: actions/setup-dotnet@v4
         with:
           dotnet-version: |
             6.0.x
-            7.0.x
 
       - name: Check dotnet
         run: dotnet --info
@@ -171,15 +125,6 @@ jobs:
           echo "----------/tmp/linux----------"
           ls -lh /tmp/linux
 
-          echo "----------/tmp/macos----------"
-          ls -lh /tmp/macos
-
-          echo "----------/tmp/windows-x64----------"
-          ls -lh /tmp/windows-x64
-
-          echo "----------/tmp/windows-x86----------"
-          ls -lh /tmp/windows-x86
-
       - name: Build
         shell: bash
         run: |

.github/workflows/test-go.yaml

@@ -127,7 +127,7 @@ jobs:
 
       - uses: actions/upload-artifact@v4
         with:
-          name: tts-waves
+          name: tts-waves-${{ matrix.os }}
           path: tts-waves
 
       - name: Test non-streaming decoding files (macOS)
@@ -154,6 +154,7 @@ jobs:
 
           echo "Test paraformer"
           ./run-paraformer.sh
+          ./run-paraformer-itn.sh
           rm -rf sherpa-onnx-paraformer-zh-2023-03-28
 
           echo "Test NeMo CTC"

.github/workflows/test-nodejs-addon-api.yaml

@@ -39,8 +39,8 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [macos-11, macos-14, ubuntu-20.04, ubuntu-22.04] #, windows-latest]
-        node-version: ["16", "17", "18", "19", "21", "22"]
+        os: [macos-latest, ubuntu-latest, ubuntu-latest]
+        node-version: ["16", "22"]
         python-version: ["3.8"]
 
     steps:

.gitignore

@@ -107,3 +107,4 @@ package-lock.json
 sherpa-onnx-nemo-*
 sherpa-onnx-vits-*
 sherpa-onnx-telespeech-ctc-*
+*.fst

dart-api-examples/non-streaming-asr/bin/paraformer-itn.dart

@@ -0,0 +1,63 @@
+// Copyright (c)  2024  Xiaomi Corporation
+import 'dart:io';
+import 'dart:typed_data';
+
+import 'package:args/args.dart';
+import 'package:sherpa_onnx/sherpa_onnx.dart' as sherpa_onnx;
+
+import './init.dart';
+
+void main(List<String> arguments) async {
+  await initSherpaOnnx();
+
+  final parser = ArgParser()
+    ..addOption('model', help: 'Path to the paraformer model')
+    ..addOption('tokens', help: 'Path to tokens.txt')
+    ..addOption('rule-fsts',
+        help: 'Path to rule fsts for inverse text normalization')
+    ..addOption('input-wav', help: 'Path to input.wav to transcribe');
+
+  final res = parser.parse(arguments);
+  if (res['model'] == null ||
+      res['tokens'] == null ||
+      res['rule-fsts'] == null ||
+      res['input-wav'] == null) {
+    print(parser.usage);
+    exit(1);
+  }
+
+  final model = res['model'] as String;
+  final tokens = res['tokens'] as String;
+  final ruleFsts = res['rule-fsts'] as String;
+  final inputWav = res['input-wav'] as String;
+
+  final paraformer = sherpa_onnx.OfflineParaformerModelConfig(
+    model: model,
+  );
+
+  final modelConfig = sherpa_onnx.OfflineModelConfig(
+    paraformer: paraformer,
+    tokens: tokens,
+    debug: true,
+    numThreads: 1,
+    modelType: 'paraformer',
+  );
+  final config = sherpa_onnx.OfflineRecognizerConfig(
+    model: modelConfig,
+    ruleFsts: ruleFsts,
+  );
+  final recognizer = sherpa_onnx.OfflineRecognizer(config);
+
+  final waveData = sherpa_onnx.readWave(inputWav);
+  final stream = recognizer.createStream();
+
+  stream.acceptWaveform(
+      samples: waveData.samples, sampleRate: waveData.sampleRate);
+  recognizer.decode(stream);
+
+  final result = recognizer.getResult(stream);
+  print(result.text);
+
+  stream.free();
+  recognizer.free();
+}

dart-api-examples/non-streaming-asr/run-paraformer-itn.sh

@@ -0,0 +1,27 @@
+#!/usr/bin/env bash
+
+set -ex
+
+dart pub get
+
+if [ ! -f ./sherpa-onnx-paraformer-zh-2023-03-28/tokens.txt ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+
+  tar xvf sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+  rm sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+fi
+
+if [ ! -f ./itn-zh-number.wav ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn-zh-number.wav
+fi
+
+if [ ! -f ./itn_zh_number.fst ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn_zh_number.fst
+fi
+
+dart run \
+  ./bin/paraformer-itn.dart \
+  --model ./sherpa-onnx-paraformer-zh-2023-03-28/model.int8.onnx \
+  --tokens ./sherpa-onnx-paraformer-zh-2023-03-28/tokens.txt \
+  --rule-fsts ./itn_zh_number.fst \
+  --input-wav ./itn-zh-number.wav

dotnet-examples/offline-decode-files/Program.cs

@@ -23,34 +23,34 @@
     public int FeatureDim { get; set; }
 
     [Option(Required = false, HelpText = "Path to tokens.txt")]
     public string Tokens { get; set; }
 
     [Option(Required = false, Default = "", HelpText = "Path to transducer encoder.onnx. Used only for transducer models")]
     public string Encoder { get; set; }
 
     [Option(Required = false, Default = "", HelpText = "Path to transducer decoder.onnx. Used only for transducer models")]
     public string Decoder { get; set; }
 
     [Option(Required = false,  Default = "",HelpText = "Path to transducer joiner.onnx. Used only for transducer models")]
     public string Joiner { get; set; }
 
     [Option("model-type", Required = false, Default = "", HelpText = "model type")]
     public string ModelType { get; set; }
 
     [Option("whisper-encoder", Required = false, Default = "", HelpText = "Path to whisper encoder.onnx. Used only for whisper models")]
     public string WhisperEncoder { get; set; }
 
     [Option("whisper-decoder", Required = false, Default = "", HelpText = "Path to whisper decoder.onnx. Used only for whisper models")]
     public string WhisperDecoder { get; set; }
 
     [Option("whisper-language", Required = false, Default = "", HelpText = "Language of the input file. Can be empty")]
     public string WhisperLanguage{ get; set; }
 
     [Option("whisper-task", Required = false, Default = "transcribe", HelpText = "transcribe or translate")]
     public string WhisperTask{ get; set; }
 
     [Option("tdnn-model", Required = false, Default = "", HelpText = "Path to tdnn yesno model")]
     public string TdnnModel { get; set; }
 
 
     [Option(Required = false, HelpText = "Path to model.onnx. Used only for paraformer models")]
@@ -69,6 +69,10 @@
             HelpText = "Valid decoding methods are: greedy_search, modified_beam_search")]
     public string DecodingMethod { get; set; }
 
+    [Option("rule-fsts", Required = false, Default = "",
+            HelpText = "If not empty, path to rule fst for inverse text normalization")]
+    public string RuleFsts { get; set; }
+
     [Option("max-active-paths", Required = false, Default = 4,
         HelpText = @"Used only when --decoding--method is modified_beam_search.
 It specifies number of active paths to keep during the search")]
@@ -233,6 +237,7 @@
     config.MaxActivePaths = options.MaxActivePaths;
     config.HotwordsFile = options.HotwordsFile;
     config.HotwordsScore = options.HotwordsScore;
+    config.RuleFsts = options.RuleFsts;
 
     config.ModelConfig.Debug = 0;
 

dotnet-examples/offline-decode-files/run-paraformer-itn.sh

@@ -0,0 +1,24 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [ ! -d ./sherpa-onnx-paraformer-zh-2023-03-28 ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+  tar xvf sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+  rm sherpa-onnx-paraformer-zh-2023-03-28.tar.bz2
+fi
+
+if [ ! -f ./itn-zh-number.wav ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn-zh-number.wav
+fi
+
+if [ ! -f ./itn_zh_number.fst ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/itn_zh_number.fst
+fi
+
+dotnet run \
+  --tokens=./sherpa-onnx-paraformer-zh-2023-03-28/tokens.txt \
+  --paraformer=./sherpa-onnx-paraformer-zh-2023-03-28/model.onnx \
+  --rule-fsts=./itn_zh_number.fst \
+  --num-threads=2 \
+  --files ./itn-zh-number.wav