|Ankit| update proto

Author: ankit-thoughtworks

examples/java/speech-recognition-client/build.gradle

@@ -21,6 +21,7 @@ protobuf {
 
 group 'org.ekstep'
 version '1.0-SNAPSHOT'
+sourceCompatibility = '1.8'
 
 repositories {
     mavenCentral()

examples/java/speech-recognition-client/src/main/java/com/ekstep/endpoints/speech_recognition/SpeechRecognitionClient.java

@@ -2,6 +2,7 @@
 
 import com.google.protobuf.ByteString;
 import io.grpc.*;
+import io.grpc.netty.shaded.io.netty.util.internal.StringUtil;
 import io.grpc.stub.MetadataUtils;
 
 import java.util.concurrent.TimeUnit;
@@ -47,25 +48,37 @@ public SpeechRecognitionResult transcribeAudioBytes(ByteString audioBytes, Langu
         }
     }
 
-    public static ByteString getAudioBytes() {
+    public static ByteString getAudioBytes(String file) {
         AudioFiles audioFiles = new AudioFiles();
-        String file = "/Users/nireshkumarr/Documents/ekstep/speech-recognition-open-api/examples/python/speech-recognition/changed.wav";
         byte[] data2 = audioFiles.readAudioFileData(file);
         ByteString byteString = ByteString.copyFrom(data2);
         return byteString;
     }
 
+    private PunctuateResponse getPunctuate(String text, String language) {
+        return blockingStub.punctuate(PunctuateRequest.newBuilder().setLanguage(language).clearEnabledItn().setText(text).build());
+    }
+
     public static void main(String[] args) throws Exception {
-        String target = "test-model-api.vakyansh.in:50051";
+        String targetGRPCServer = "<GRPC Server>";
+        String audioFile = "<wav audio file path>";
+        Language.LanguageCode languageCode = Language.LanguageCode.hi;
 
-        ManagedChannel channel = ManagedChannelBuilder.forTarget(target)
+        ManagedChannel channel = ManagedChannelBuilder.forTarget(targetGRPCServer)
                 .usePlaintext()
                 .build();
         try {
             SpeechRecognitionClient client = new SpeechRecognitionClient(channel);
-            ByteString audioBytes = getAudioBytes();
-            SpeechRecognitionResult result = client.transcribeAudioBytes(audioBytes, Language.LanguageCode.en, RecognitionConfig.AudioFormat.wav, RecognitionConfig.TranscriptionFormatEnum.transcript);
-            System.out.println(result.toString());
+            ByteString audioBytes = getAudioBytes(audioFile);
+            SpeechRecognitionResult result = client.transcribeAudioBytes(audioBytes, languageCode, RecognitionConfig.AudioFormat.wav, RecognitionConfig.TranscriptionFormatEnum.transcript);
+            System.out.println("#### Output ####");
+            for (SpeechRecognitionResult.Output out : result.getOutputList()) {
+                if (!StringUtil.isNullOrEmpty(out.getSource())) {
+                    System.out.println("Text: " + out.getSource());
+                    System.out.println("Punctuated Text: " + client.getPunctuate(out.getSource(), languageCode.name()).getText());
+                }
+            }
+
 
         } finally {
             channel.shutdownNow().awaitTermination(5, TimeUnit.SECONDS);

examples/java/speech-recognition-client/src/main/proto/speech-recognition-open-api.proto

@@ -9,11 +9,31 @@ option java_package = "com.ekstep.endpoints.speech_recognition";
 import "google/api/annotations.proto";
 
 service SpeechRecognizer {
+  rpc recognize_audio(stream Message) returns (stream Response) {}
+  rpc punctuate(PunctuateRequest) returns (PunctuateResponse) {
+    option (google.api.http) = {post: "/v1/punctuate/{language}", body: "*"};
+  }
   rpc recognize (SpeechRecognitionRequest) returns (SpeechRecognitionResult) {
     option (google.api.http) = {post: "/v1/recognize/{config.language.sourceLanguage}", body: "*"};
   }
 }
 
+message Message {
+  bytes audio = 1;
+  string user = 2;
+  string language = 3;
+  bool speaking = 4;
+  bool isEnd = 5;
+}
+
+message Response {
+  string transcription = 1;
+  string user = 2;
+  string language = 3;
+  string action = 4;
+}
+
+
 message SpeechRecognitionRequest {
   RecognitionConfig config = 1;
   repeated RecognitionAudio audio = 2;
@@ -107,7 +127,9 @@ message Language {
     grt = 28;
     sa = 29;
     raj = 30;
-    bo = 31;
+    bho = 31;
+    en_bio = 32;
+    hi_en = 33;
   }
   optional string name = 1;
   LanguageCode sourceLanguage = 2;
@@ -148,3 +170,14 @@ message Alternative {
   string startTime = 2;
   string endTime = 3;
 }
+
+message PunctuateRequest {
+  string text = 1;
+  string language = 2;
+  bool enabledItn = 3;
+}
+
+message PunctuateResponse {
+  string text = 1;
+  string language = 2;
+}