Add Go API for homophone replacer (#2168)

.github/workflows/test-go.yaml

@@ -142,32 +142,31 @@ jobs:
           name: ${{ matrix.os }}-libs
           path: to-upload/
 
-      - name: Test speech enhancement (GTCRN)
+      - name: Test streaming decoding files
         shell: bash
         run: |
-          cd scripts/go/_internal/speech-enhancement-gtcrn/
+          cd scripts/go/_internal/streaming-decode-files
-
+          ls -lh
-          ./run.sh
+          go mod tidy
-
+          cat go.mod
+          go build
           ls -lh
 
-      - name: Test audio tagging
+          echo "Test zipformer2 CTC"
-        shell: bash
+          ./run-zipformer2-ctc-with-hr.sh
-        run: |
+          ./run-zipformer2-ctc.sh
-          cd scripts/go/_internal/audio-tagging/
+          rm -rf sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13
 
-          ./run.sh
+          echo "Test transducer"
+          ./run-transducer.sh
+          rm -rf sherpa-onnx-streaming-zipformer-en-2023-06-26
 
-          ls -lh
+          ./run-transducer-itn.sh
+          rm -rf sherpa-onnx-streaming-*
 
-      - name: Test Keyword spotting
+          echo "Test paraformer"
-        shell: bash
+          ./run-paraformer.sh
-        run: |
+          rm -rf sherpa-onnx-streaming-paraformer-bilingual-zh-en
-          cd scripts/go/_internal/keyword-spotting-from-file/
-
-          ./run.sh
-
-          ls -lh
 
       - name: Test non-streaming decoding files
         shell: bash
@@ -179,6 +178,11 @@ jobs:
           go build
           ls -lh
 
+          echo "Test SenseVoice ctc"
+          ./run-sense-voice-small-with-hr.sh
+          ./run-sense-voice-small.sh
+          rm -rf sherpa-onnx-sense-*
+
           echo "Test Dolphin CTC"
           ./run-dolphin-ctc-base.sh
           rm -rf sherpa-onnx-dolphin-*
@@ -191,10 +195,6 @@ jobs:
           ./run-moonshine.sh
           rm -rf sherpa-onnx-*
 
-          echo "Test SenseVoice ctc"
-          ./run-sense-voice-small.sh
-          rm -rf sherpa-onnx-sense-*
-
           echo "Test telespeech ctc"
           ./run-telespeech-ctc.sh
           rm -rf sherpa-onnx-telespeech-ctc-*
@@ -224,6 +224,33 @@ jobs:
           ./run-tdnn-yesno.sh
           rm -rf sherpa-onnx-tdnn-yesno
 
+      - name: Test speech enhancement (GTCRN)
+        shell: bash
+        run: |
+          cd scripts/go/_internal/speech-enhancement-gtcrn/
+
+          ./run.sh
+
+          ls -lh
+
+      - name: Test audio tagging
+        shell: bash
+        run: |
+          cd scripts/go/_internal/audio-tagging/
+
+          ./run.sh
+
+          ls -lh
+
+      - name: Test Keyword spotting
+        shell: bash
+        run: |
+          cd scripts/go/_internal/keyword-spotting-from-file/
+
+          ./run.sh
+
+          ls -lh
+
       - name: Test adding punctuation
         shell: bash
         run: |
@@ -301,28 +328,3 @@ jobs:
         with:
           name: tts-waves-${{ matrix.os }}
           path: tts-waves
-
-      - name: Test streaming decoding files
-        shell: bash
-        run: |
-          cd scripts/go/_internal/streaming-decode-files
-          ls -lh
-          go mod tidy
-          cat go.mod
-          go build
-          ls -lh
-
-          echo "Test zipformer2 CTC"
-          ./run-zipformer2-ctc.sh
-          rm -rf sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13
-
-          echo "Test transducer"
-          ./run-transducer.sh
-          rm -rf sherpa-onnx-streaming-zipformer-en-2023-06-26
-
-          ./run-transducer-itn.sh
-          rm -rf sherpa-onnx-streaming-*
-
-          echo "Test paraformer"
-          ./run-paraformer.sh
-          rm -rf sherpa-onnx-streaming-paraformer-bilingual-zh-en

go-api-examples/non-streaming-decode-files/main.go

@@ -66,6 +66,10 @@ func main() {
 	flag.StringVar(&config.RuleFsts, "rule-fsts", "", "If not empty, path to rule fst for inverse text normalization")
 	flag.StringVar(&config.RuleFars, "rule-fars", "", "If not empty, path to rule fst archives for inverse text normalization")
 
+	flag.StringVar(&config.Hr.DictDir, "hr-dict-dir", "", "If not empty, path to the jieba dict dir for homonphone replacer")
+	flag.StringVar(&config.Hr.Lexicon, "hr-lexicon", "", "If not empty, path to the lexicon.txt for homonphone replacer")
+	flag.StringVar(&config.Hr.RuleFsts, "hr-rule-fsts", "", "If not empty, path to the replace.fst for homonphone replacer")
+
 	flag.Parse()
 
 	if len(flag.Args()) != 1 {

go-api-examples/non-streaming-decode-files/run-sense-voice-small-with-hr.sh

@@ -0,0 +1,31 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [ ! -d sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17  ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
+  tar xvf sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
+  rm sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
+fi
+
+if [ ! -d dict ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/dict.tar.bz2
+  tar xf dict.tar.bz2
+  rm dict.tar.bz2
+
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/replace.fst
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/test-hr.wav
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/lexicon.txt
+fi
+
+go mod tidy
+go build
+
+./non-streaming-decode-files \
+  --sense-voice-model ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/model.int8.onnx \
+  --tokens ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tokens.txt \
+  --debug 1 \
+  --hr-dict-dir ./dict \
+  --hr-lexicon ./lexicon.txt \
+  --hr-rule-fsts ./replace.fst \
+  ./test-hr.wav

go-api-examples/streaming-decode-files/main.go

@@ -32,6 +32,9 @@ func main() {
 	flag.IntVar(&config.MaxActivePaths, "max-active-paths", 4, "Used only when --decoding-method is modified_beam_search")
 	flag.StringVar(&config.RuleFsts, "rule-fsts", "", "If not empty, path to rule fst for inverse text normalization")
 	flag.StringVar(&config.RuleFars, "rule-fars", "", "If not empty, path to rule fst archives for inverse text normalization")
+	flag.StringVar(&config.Hr.DictDir, "hr-dict-dir", "", "If not empty, path to the jieba dict dir for homonphone replacer")
+	flag.StringVar(&config.Hr.Lexicon, "hr-lexicon", "", "If not empty, path to the lexicon.txt for homonphone replacer")
+	flag.StringVar(&config.Hr.RuleFsts, "hr-rule-fsts", "", "If not empty, path to the replace.fst for homonphone replacer")
 
 	flag.Parse()
 

go-api-examples/streaming-decode-files/run-zipformer2-ctc-with-hr.sh

@@ -0,0 +1,30 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [ ! -d sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13 ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13.tar.bz2
+  tar xvf sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13.tar.bz2
+  rm sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13.tar.bz2
+fi
+
+if [ ! -d dict ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/dict.tar.bz2
+  tar xf dict.tar.bz2
+  rm dict.tar.bz2
+
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/replace.fst
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/test-hr.wav
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/hr-files/lexicon.txt
+fi
+
+go mod tidy
+go build
+
+./streaming-decode-files \
+  --zipformer2-ctc ./sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13/ctc-epoch-20-avg-1-chunk-16-left-128.onnx \
+  --tokens ./sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13/tokens.txt \
+  --hr-dict-dir ./dict \
+  --hr-lexicon ./lexicon.txt \
+  --hr-rule-fsts ./replace.fst \
+  ./test-hr.wav

scripts/go/_internal/non-streaming-decode-files/run-sense-voice-small-with-hr.sh

@@ -0,0 +1 @@
+../../../../go-api-examples/non-streaming-decode-files/run-sense-voice-small-with-hr.sh

scripts/go/_internal/streaming-decode-files/run-zipformer2-ctc-with-hr.sh

@@ -0,0 +1 @@
+../../../../go-api-examples/streaming-decode-files/run-zipformer2-ctc-with-hr.sh

scripts/go/sherpa_onnx.go

@@ -108,6 +108,12 @@ type OnlineCtcFstDecoderConfig struct {
 	MaxActive int
 }
 
+type HomophoneReplacerConfig struct {
+	DictDir  string
+	Lexicon  string
+	RuleFsts string
+}
+
 // Configuration for the online/streaming recognizer.
 type OnlineRecognizerConfig struct {
 	FeatConfig  FeatureConfig
@@ -137,6 +143,7 @@ type OnlineRecognizerConfig struct {
 	RuleFars                string
 	HotwordsBuf             string
 	HotwordsBufSize         int
+	Hr                      HomophoneReplacerConfig
 }
 
 // It contains the recognition result for a online stream.
@@ -239,6 +246,15 @@ func NewOnlineRecognizer(config *OnlineRecognizerConfig) *OnlineRecognizer {
 	defer C.free(unsafe.Pointer(c.ctc_fst_decoder_config.graph))
 	c.ctc_fst_decoder_config.max_active = C.int(config.CtcFstDecoderConfig.MaxActive)
 
+	c.hr.dict_dir = C.CString(config.Hr.DictDir)
+	defer C.free(unsafe.Pointer(c.hr.dict_dir))
+
+	c.hr.lexicon = C.CString(config.Hr.Lexicon)
+	defer C.free(unsafe.Pointer(c.hr.lexicon))
+
+	c.hr.rule_fsts = C.CString(config.Hr.RuleFsts)
+	defer C.free(unsafe.Pointer(c.hr.rule_fsts))
+
 	impl := C.SherpaOnnxCreateOnlineRecognizer(&c)
 	if impl == nil {
 		return nil
@@ -462,6 +478,7 @@ type OfflineRecognizerConfig struct {
 	BlankPenalty   float32
 	RuleFsts       string
 	RuleFars       string
+	Hr             HomophoneReplacerConfig
 }
 
 // It wraps a pointer from C
@@ -549,6 +566,10 @@ func newCOfflineRecognizerConfig(config *OfflineRecognizerConfig) *C.struct_Sher
 
 	c.rule_fsts = C.CString(config.RuleFsts)
 	c.rule_fars = C.CString(config.RuleFars)
+
+	c.hr.dict_dir = C.CString(config.Hr.DictDir)
+	c.hr.lexicon = C.CString(config.Hr.Lexicon)
+	c.hr.rule_fsts = C.CString(config.Hr.RuleFsts)
 	return &c
 }
 func freeCOfflineRecognizerConfig(c *C.struct_SherpaOnnxOfflineRecognizerConfig) {
@@ -676,10 +697,26 @@ func freeCOfflineRecognizerConfig(c *C.struct_SherpaOnnxOfflineRecognizerConfig)
 		C.free(unsafe.Pointer(c.rule_fsts))
 		c.rule_fsts = nil
 	}
+
 	if c.rule_fars != nil {
 		C.free(unsafe.Pointer(c.rule_fars))
 		c.rule_fars = nil
 	}
+
+	if c.hr.dict_dir != nil {
+		C.free(unsafe.Pointer(c.hr.dict_dir))
+		c.hr.dict_dir = nil
+	}
+
+	if c.hr.lexicon != nil {
+		C.free(unsafe.Pointer(c.hr.lexicon))
+		c.hr.lexicon = nil
+	}
+
+	if c.hr.rule_fsts != nil {
+		C.free(unsafe.Pointer(c.hr.rule_fsts))
+		c.hr.rule_fsts = nil
+	}
 }
 
 // Frees the internal pointer of the recognition to avoid memory leak.