AISDK-229: Add skip_postprocessing option (#56)

Author: Kyle Bridburg

examples/AsyncTranscribeLocalMediaFile.java

@@ -32,6 +32,7 @@ public static void main(String[] args) {
     revAiJobOptions.setNotificationConfig("https://example.com");
     revAiJobOptions.setSkipPunctuation(false);
     revAiJobOptions.setSkipDiarization(false);
+    revAiJobOptions.setSkipPostprocessing(false);
     revAiJobOptions.setFilterProfanity(true);
     revAiJobOptions.setRemoveDisfluencies(true);
     revAiJobOptions.setSpeakerChannelsCount(null);

examples/AsyncTranscribeMediaUrl.java

@@ -47,6 +47,7 @@ public static void main(String[] args) {
     revAiJobOptions.setNotificationConfig(callbackUrl, notificationAuth);
     revAiJobOptions.setSkipPunctuation(false);
     revAiJobOptions.setSkipDiarization(false);
+    revAiJobOptions.setSkipPostprocessing(false);
     revAiJobOptions.setFilterProfanity(true);
     revAiJobOptions.setRemoveDisfluencies(true);
     revAiJobOptions.setSpeakerChannelsCount(null);

examples/StreamingFromLocalFile.java

@@ -39,6 +39,7 @@ public static void main(String[] args) throws InterruptedException {
     sessionConfig.setStartTs(0.0);
     sessionConfig.setTranscriber("machine");
     sessionConfig.setLanguage("en");
+    sessionConfig.setSkipPostprocessing(false);
 
     // Initialize your client with your access token
     StreamingClient streamingClient = new StreamingClient(accessToken);

src/main/java/ai/rev/speechtotext/StreamingClient.java

@@ -217,6 +217,10 @@ private String buildURL(
       if (sessionConfig.getLanguage() != null) {
         urlBuilder.addQueryParameter("language", sessionConfig.getLanguage());
       }
+      if (sessionConfig.getSkipPostprocessing() != null) {
+        urlBuilder.addQueryParameter(
+          "skip_postprocessing", String.valueOf(sessionConfig.getSkipPostprocessing()));
+      }
     }
     return urlBuilder.build().toString()
       + "&content_type="

src/main/java/ai/rev/speechtotext/models/asynchronous/RevAiJobOptions.java

@@ -44,6 +44,10 @@ public class RevAiJobOptions {
   /** Optional parameter for the speech engine to skip punctuation. */
   @SerializedName("skip_punctuation")
   private Boolean skipPunctuation;
+  
+  /** Optional parameter for the speech engine to skip all postprocessing */
+  @SerializedName("skip_postprocessing")
+  private Boolean skipPostprocessing;
 
   /**
    * Optional parameter to process each audio channel separately. Account will be charged the file
@@ -273,6 +277,25 @@ public Boolean getSkipPunctuation() {
   public void setSkipPunctuation(Boolean skipPunctuation) {
     this.skipPunctuation = skipPunctuation;
   }
+  
+  /**
+   * Returns the value of the skip postprocessing Boolean.
+   *
+   * @return The skip postprocessing value.
+   */
+  public Boolean getSkipPostprocessing() {
+    return skipPostprocessing;
+  }
+
+  /**
+   * Specifies if all postprocessing (capitalization, punctuation, ITN) will be skipped by the speech engine. This property is
+   * optional and defaults to false.
+   *
+   * @param skipPostprocessing The value of the Boolean.
+   */
+  public void setSkipPostprocessing(Boolean skipPostprocessing) {
+    this.skipPostprocessing = skipPostprocessing;
+  }
 
   /**
    * Returns the speaker channel count.

src/main/java/ai/rev/speechtotext/models/streaming/SessionConfig.java

@@ -12,6 +12,7 @@ public class SessionConfig {
   private Double startTs;
   private String transcriber;
   private String language;
+  private Boolean skipPostprocessing;
 
   /**
    * Returns the metadata.
@@ -138,40 +139,71 @@ public void setDetailedPartials(Boolean detailedPartials) {
    *
    * @return The startTs value.
    */
-  public Double getStartTs() { return startTs; }
+  public Double getStartTs() {
+    return startTs;
+  }
 
   /**
    * Specifies the number of seconds to offset all hypotheses timings.
    *
    * @param startTs The number of seconds to offset all hypotheses timings.
    */
-  public void setStartTs(Double startTs) { this.startTs = startTs; }
+  public void setStartTs(Double startTs) {
+    this.startTs = startTs;
+  }
 
   /**
    * Returns the value of transcriber.
    *
    * @return The transcriber value.
    */
-  public String getTranscriber() { return transcriber; }
+  public String getTranscriber() {
+    return transcriber;
+  }
 
   /**
    * Specifies the type of transcriber to use to transcribe the media.
    *
    * @param transcriber The type of transcriber to use to transcribe the media.
    */
-  public void setTranscriber(String transcriber) { this.transcriber = transcriber; }
+  public void setTranscriber(String transcriber) {
+    this.transcriber = transcriber;
+  }
 
   /**
    * Returns the value of language.
    *
    * @return The language value.
    */
-  public String getLanguage() { return language; }
+  public String getLanguage() {
+    return language;
+  }
 
   /**
    * Specifies the language to use for the streaming job.
    *
    * @param language The language to use for the streaming job.
    */
-  public void setLanguage(String language) { this.language = language; }
+  public void setLanguage(String language) {
+    this.language = language;
+  }
+  
+  /**
+   * Returns the value of the skip postprocessing Boolean.
+   *
+   * @return The skip postprocessing value.
+   */
+  public Boolean getSkipPostprocessing() {
+    return skipPostprocessing;
+  }
+
+  /**
+   * Specifies if all postprocessing (capitalization, punctuation, ITN) will be skipped by the speech engine. This property is
+   * optional and defaults to false.
+   *
+   * @param skipPostprocessing The value of the Boolean.
+   */
+  public void setSkipPostprocessing(Boolean skipPostprocessing) {
+    this.skipPostprocessing = skipPostprocessing;
+  }
 }

src/test/java/ai/rev/speechtotext/integration/StreamingTest.java

@@ -132,8 +132,9 @@ private SessionConfig createSessionsConfig() {
     sessionConfig.setDeleteAfterSeconds(0);
     sessionConfig.setDetailedPartials(true);
     sessionConfig.setStartTs(10.0);
-    sessionConfig.setTranscriber("machine");
+    sessionConfig.setTranscriber("machine_v2");
     sessionConfig.setLanguage("en");
+    sessionConfig.setSkipPostprocessing(false);
     return sessionConfig;
   }
 

src/test/java/ai/rev/speechtotext/integration/SubmitJobTest.java

@@ -171,6 +171,7 @@ private RevAiJobOptions getJobOptions() {
     revAiJobOptions.setRemoveDisfluencies(true);
     revAiJobOptions.setSkipPunctuation(true);
     revAiJobOptions.setSkipDiarization(true);
+    revAiJobOptions.setSkipPostprocessing(true);
     revAiJobOptions.setSpeakerChannelsCount(null);
     revAiJobOptions.setDeleteAfterSeconds(0);
     revAiJobOptions.setLanguage("en");

src/test/java/ai/rev/speechtotext/unit/RevAiJobTest.java

@@ -159,6 +159,7 @@ public void SubmitJobUrl_UrlAndOptionsAreSpecified_ReturnsARevAiJob() throws IOE
     options.setRemoveDisfluencies(true);
     options.setSkipPunctuation(true);
     options.setSkipDiarization(true);
+    options.setSkipPostprocessing(true);
     options.setNotificationConfig("https://example.com");
     options.setMetadata(METADATA);
     options.setSpeakerChannelsCount(2);
@@ -196,6 +197,7 @@ public void SubmitJobUrl_UrlAndOptionsAreSpecified_WithAuthHeaders_ReturnsARevAi
     options.setRemoveDisfluencies(true);
     options.setSkipPunctuation(true);
     options.setSkipDiarization(true);
+    options.setSkipPostprocessing(true);
     options.setNotificationConfig(NOTIFICATION_URL, NOTIFICATION_AUTH);
     options.setMetadata(METADATA);
     options.setSpeakerChannelsCount(2);

src/test/java/ai/rev/speechtotext/unit/RevAiStreamingClientOptionalParametersTest.java

@@ -48,17 +48,17 @@ public static String[] vocabIdAndNull() {
 
   @DataPoints("deleteAfterSecondsAndNull")
   public static Integer[] deleteAfterSecondsAndNull() {
-    return new Integer[] {0, 1, null};
+    return new Integer[] {1, null};
   }
 
   @DataPoints("booleanValuesAndNull")
   public static Boolean[] booleanValuesAndNull() {
-    return new Boolean[] {true, false, null};
+    return new Boolean[] {false, null};
   }
 
   @DataPoints("startTsAndNull")
   public static Double[] startTsAndNull() {
-    return new Double[] {0.0, 10.0, null};
+    return new Double[] {10.0, null};
   }
 
   @DataPoints("transcriberAndNull")
@@ -114,7 +114,8 @@ public void StreamingClient_GivenOptionalParameterSet_ContainsNonNullParametersI
       @FromDataPoints("booleanValuesAndNull") Boolean detailedPartials,
       @FromDataPoints("startTsAndNull") Double startTs,
       @FromDataPoints("transcriberAndNull") String transcriber,
-      @FromDataPoints("languageAndNull") String language)
+      @FromDataPoints("languageAndNull") String language,
+      @FromDataPoints("booleanValuesAndNull") Boolean skipPostprocessing)
       throws UnsupportedEncodingException {
     sessionConfig.setMetaData(metadata);
     sessionConfig.setFilterProfanity(filterProfanity);
@@ -125,6 +126,7 @@ public void StreamingClient_GivenOptionalParameterSet_ContainsNonNullParametersI
     sessionConfig.setStartTs(startTs);
     sessionConfig.setTranscriber(transcriber);
     sessionConfig.setLanguage(language);
+    sessionConfig.setSkipPostprocessing(skipPostprocessing);
 
     streamingClient.connect(
         Mockito.mock(RevAiWebSocketListener.class), defaultContentType, sessionConfig);
@@ -179,6 +181,10 @@ private void assertStreamOptions(RecordedRequest request) throws UnsupportedEnco
       assertThat(request.getPath())
         .contains("language=" + sessionConfig.getLanguage());
     }
+    if (sessionConfig.getSkipPostprocessing() != null) {
+      assertThat(request.getPath())
+        .contains("skip_postprocessing=" + sessionConfig.getSkipPostprocessing());
+    }
     assertThat(request.getPath()).contains("access_token=" + FAKE_ACCESS_TOKEN);
   }
 }