Configurable Google Speech to Text transcription

README.md

@@ -126,6 +126,15 @@ sudo apt-get update && sudo apt-get install google-cloud-sdk google-cloud-sdk-ap
 gcloud init
 gcloud auth application-default login
 ```
+It is possible to enable or disable the functionality of Google Cloud Speech to Text.
+By default, the properties
+`org.jitsi.jigasi.transcription.ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION=false`
+and
+`org.jitsi.jigasi.transcription.ENABLE_GOOGLE_PROFANITY_FILTER=false`
+in
+`jigasi-home/sip-communicator.properties`
+disable automatic punctuation, profanity filter results for the transcription.
+To change this, simply set the desired property to `true` or `false`.
 
 Vosk configuration
 ==================
@@ -206,6 +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)

src/main/java/org/jitsi/jigasi/transcription/GoogleCloudTranscriptionService.java

@@ -184,11 +184,35 @@ public class GoogleCloudTranscriptionService
     private final static String P_NAME_USE_VIDEO_MODEL
         = "org.jitsi.jigasi.transcription.USE_VIDEO_MODEL";
 
+    /**
+     * Property name to determine whether the Google Speech API should get
+     * automatic punctuation
+     */
+    private final static String P_NAME_ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION
+        = "org.jitsi.jigasi.transcription.ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION";
+
+    /**
+     * Property name to determine whether the Google Speech API should censor
+     * profane words
+     */
+    private final static String P_NAME_ENABLE_GOOGLE_PROFANITY_FILTER
+        = "org.jitsi.jigasi.transcription.ENABLE_GOOGLE_PROFANITY_FILTER";
+
     /**
      * The default value for the property USE_VIDEO_MODEL
      */
     private final static boolean DEFAULT_VALUE_USE_VIDEO_MODEL = false;
 
+    /**
+     * The default value for the property ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION
+     */
+    private final static boolean DEFAULT_VALUE_ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION = false;
+
+    /**
+     * The default value for the property ENABLE_GOOGLE_PROFANITY_FILTER
+     */
+    private final static boolean DEFAULT_VALUE_ENABLE_GOOGLE_PROFANITY_FILTER = false;
+
     /**
      * Check whether the given string contains a supported language tag
      *
@@ -224,6 +248,16 @@ private static void validateLanguageTag(String tag)
      */
     private boolean useVideoModel;
 
+    /**
+     * Whether to get automatic punctuation
+     */
+    private boolean enableAutomaticPunctuation;
+
+    /**
+     * Wheteher to enable profanity filter
+     */
+    private boolean enableProfanityFilter;
+
     /**
      * Creates the RecognitionConfig the Google service uses based
      * on the TranscriptionRequest
@@ -263,6 +297,12 @@ private RecognitionConfig getRecognitionConfig(TranscriptionRequest request)
             builder.setModel("video");
         }
 
+        // set punctuation mode
+        builder.setEnableAutomaticPunctuation(enableAutomaticPunctuation);
+
+        // set profanity filter
+        builder.setProfanityFilter(enableProfanityFilter);
+
         // set the Language tag
         String languageTag = request.getLocale().toLanguageTag();
         validateLanguageTag(languageTag);
@@ -284,6 +324,12 @@ public GoogleCloudTranscriptionService()
     {
         useVideoModel = JigasiBundleActivator.getConfigurationService()
             .getBoolean(P_NAME_USE_VIDEO_MODEL, DEFAULT_VALUE_USE_VIDEO_MODEL);
+
+        enableAutomaticPunctuation = JigasiBundleActivator.getConfigurationService()
+            .getBoolean(P_NAME_ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION, DEFAULT_VALUE_ENABLE_GOOGLE_AUTOMATIC_PUNCTUATION);
+
+        enableProfanityFilter = JigasiBundleActivator.getConfigurationService()
+            .getBoolean(P_NAME_ENABLE_GOOGLE_PROFANITY_FILTER, DEFAULT_VALUE_ENABLE_GOOGLE_PROFANITY_FILTER);
     }
 
     /**

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);