Moved ENABLE_INTERIM_RESULTS parameter from GoogleCloudTranscriptionS…

…ervice to RemotePublisherTranscriptionHandler
jitsi · Jun 14, 2022 · 886a387 · 886a387
1 parent 40d7fb2
commit 886a387
Show file tree

Hide file tree

Showing 3 changed files with 30 additions and 34 deletions.
diff --git a/README.md b/README.md
@@ -131,11 +131,9 @@ By default, the properties
 `org.jitsi.jigasi.transcription.ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION=false`
 and
 `org.jitsi.jigasi.transcription.ENABLE_GOOGLE_PROFANITY_FILTER=false`
-and
-`org.jitsi.jigasi.transcription.ENABLE_GOOGLE_INTERIM_RESULTS=false`
 in
 `jigasi-home/sip-communicator.properties`
-disable automatic punctuation, profanity filter and interim results for the transcription.
+disable automatic punctuation, profanity filter results for the transcription.
 To change this, simply set the desired property to `true` or `false`.
 
 Vosk configuration
@@ -217,6 +215,11 @@ XMPP account must also be set to make Jigasi be able to join a conference room.
             in plain text. Note that this will result in the chat being somewhat
             spammed.</td>
     </tr>
+    <tr>
+        <td>org.jitsi.jigasi.transcription.ENABLE_INTERIM_RESULTS</td>
+        <td>false</td>
+        <td>Whether or not to send interim non-final results. Note that interim results should be handled so that no repeated transcriptions are displayed to the user.</td>
+    </tr>
 </table>
 
 Call control MUCs (brewery)

diff --git a/src/main/java/org/jitsi/jigasi/transcription/GoogleCloudTranscriptionService.java b/src/main/java/org/jitsi/jigasi/transcription/GoogleCloudTranscriptionService.java
@@ -184,12 +184,6 @@ public class GoogleCloudTranscriptionService
     private final static String P_NAME_USE_VIDEO_MODEL
         = "org.jitsi.jigasi.transcription.USE_VIDEO_MODEL";
 
-    /**
-     * Property name to determine whether to send the interim results
-     */
-    private final static String P_NAME_ENABLE_GOOGLE_INTERIM_RESULTS
-        = "org.jitsi.jigasi.transcription.ENABLE_GOOGLE_INTERIM_RESULTS";
-
     /**
      * Property name to determine whether the Google Speech API should get
      * automatic punctuation
@@ -209,11 +203,6 @@ public class GoogleCloudTranscriptionService
      */
     private final static boolean DEFAULT_VALUE_USE_VIDEO_MODEL = false;
 
-    /**
-     * The default value for the property ENABLE_GOOGLE_INTERIM_RESULTS
-     */
-    private final static boolean DEFAULT_VALUE_ENABLE_GOOGLE_INTERIM_RESULTS = false;
-
     /**
      * The default value for the property ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION
      */
@@ -259,11 +248,6 @@ private static void validateLanguageTag(String tag)
      */
     private boolean useVideoModel;
 
-    /**
-     * Whether to send interim non-final results
-     */
-    private boolean enableInterimResults;
-
     /**
      * Whether to get automatic punctuation
      */
@@ -341,9 +325,6 @@ public GoogleCloudTranscriptionService()
         useVideoModel = JigasiBundleActivator.getConfigurationService()
             .getBoolean(P_NAME_USE_VIDEO_MODEL, DEFAULT_VALUE_USE_VIDEO_MODEL);
 
-        enableInterimResults = JigasiBundleActivator.getConfigurationService()
-            .getBoolean(P_NAME_ENABLE_GOOGLE_INTERIM_RESULTS, DEFAULT_VALUE_ENABLE_GOOGLE_INTERIM_RESULTS);
-
         enableAutomaticPunctuation = JigasiBundleActivator.getConfigurationService()
             .getBoolean(P_NAME_ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION, DEFAULT_VALUE_ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION);
 
@@ -745,8 +726,7 @@ private ApiStreamObserver<StreamingRecognizeRequest> createObserver(
                 new ResponseApiStreamingObserver<StreamingRecognizeResponse>(
                     this,
                     config.getLanguageCode(),
-                    debugName,
-                    enableInterimResults);
+                    debugName);
 
             // StreamingRecognitionConfig which will hold information
             // about the streaming session, including the RecognitionConfig
@@ -927,11 +907,6 @@ private static class ResponseApiStreamingObserver
          */
         private UUID messageID;
 
-        /**
-         * Whether to send interim results
-         */
-        private Boolean enableInterimResults;
-
         /**
          * Google provides multiple results per API response where the first one
          * contains the most stable part of the sentence and freshly transcribed
@@ -951,13 +926,11 @@ private static class ResponseApiStreamingObserver
          */
         ResponseApiStreamingObserver(RequestApiStreamObserverManager manager,
                                      String languageTag,
-                                     String debugName,
-                                     Boolean enableInterimResults)
+                                     String debugName)
         {
             this.requestManager = manager;
             this.languageTag = languageTag;
             this.debugName = debugName;
-            this.enableInterimResults = enableInterimResults;
 
             messageID = UUID.randomUUID();
         }
@@ -1091,7 +1064,7 @@ private void handleResult(StreamingRecognitionResult result)
             TranscriptionResult transcriptionResult = new TranscriptionResult(
                 null,
                 this.messageID,
-                !enableInterimResults && !result.getIsFinal(),
+                !result.getIsFinal(),
                 this.languageTag,
                 result.getStability(),
                 new TranscriptionAlternative(

diff --git a/src/main/java/org/jitsi/jigasi/transcription/RemotePublisherTranscriptionHandler.java b/src/main/java/org/jitsi/jigasi/transcription/RemotePublisherTranscriptionHandler.java
@@ -17,6 +17,7 @@
  */
 package org.jitsi.jigasi.transcription;
 
+import org.jitsi.jigasi.*;
 import net.java.sip.communicator.service.protocol.*;
 import org.json.*;
 
@@ -31,11 +32,27 @@ public class RemotePublisherTranscriptionHandler
     extends LocalJsonTranscriptHandler
     implements TranscriptionEventListener
 {
+    /**
+     * Property name to determine whether to send the interim results
+     */
+    private final static String P_NAME_ENABLE_INTERIM_RESULTS
+        = "org.jitsi.jigasi.transcription.ENABLE_INTERIM_RESULTS";
+
+    /**
+     * The default value for the property ENABLE_INTERIM_RESULTS
+     */
+    private final static boolean DEFAULT_VALUE_ENABLE_INTERIM_RESULTS = false;
+
     /**
      * List of remote services to notify for transcriptions.
      */
     private List<String> urls = new ArrayList<>();
 
+    /**
+     * Whether to send interim non-final results
+     */
+    private boolean enableInterimResults;
+
     /**
      * Constructs RemotePublisherTranscriptionHandler, initializing its config.
      *
@@ -52,12 +69,15 @@ public RemotePublisherTranscriptionHandler(String urlsStr)
         {
             urls.add(tokens.nextToken().trim());
         }
+
+        enableInterimResults = JigasiBundleActivator.getConfigurationService()
+            .getBoolean(P_NAME_ENABLE_INTERIM_RESULTS, DEFAULT_VALUE_ENABLE_INTERIM_RESULTS);
     }
 
     @Override
     public void publish(ChatRoom room, TranscriptionResult result)
     {
-        if (result.isInterim())
+        if (!enableInterimResults && result.isInterim())
             return;
 
         JSONObject eventObject = createTranscriptionJSONObject(result);