AISDK-202: Add Speaker Names option (#55)

Author: beaudrychase

README.md

@@ -88,6 +88,11 @@ segment.setStartTimestamp(2.0);
 segment.setEndTimestamp(100.5);
 options.setSegmentsToTranscribe(List.of(segment));
 
+// optional speaker names 
+SpeakerName speaker = new SpeakerName();
+speaker.setDisplayName('Alan Mathison Turing');
+options.setSpeakerNames(List.of(speaker));
+
 RevAiJob revAiJob = apiClient.submitJobUrl(urlLinkToFile, options);
 ```
 

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

@@ -12,6 +12,9 @@ public class Monologue {
   @SerializedName("speaker")
   private Integer speaker;
 
+  @SerializedName("speaker_info")
+  private SpeakerInfo speakerInfo;
+
   @SerializedName("elements")
   private List<Element> elements;
 
@@ -33,6 +36,24 @@ public void setSpeaker(Integer speaker) {
     this.speaker = speaker;
   }
 
+  /**
+   * Returns the speaker info for this monologue.
+   *
+   * @return The speaker info for this monologue.
+   */
+  public SpeakerInfo getSpeakerInfo() {
+    return speakerInfo;
+  }
+
+  /**
+   * Sets the speaker info for this monologue.
+   *
+   * @param speakerInfo Info about the speaker for this monologue.
+   */
+  public void setSpeakerInfo(SpeakerInfo speakerInfo) {
+    this.speakerInfo = speakerInfo;
+  }
+
   /**
    * Returns a list of {@link Element} objects.
    *

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

@@ -125,6 +125,13 @@ public class RevAiJobOptions {
   @SerializedName("segments_to_transcribe")
   private List<SegmentToTranscribe> segmentsToTranscribe;
 
+  /**
+   * Optional and only available with transcriber "human".
+   * Specifies a list of names for the speakers in an audio file.
+   */
+  @SerializedName("speaker_names")
+  private List<SpeakerName> speakerNames;
+
   /**
    * Returns the media url.
    *
@@ -514,4 +521,23 @@ public List<SegmentToTranscribe> getSegmentsToTranscribe() {
   public void setSegmentsToTranscribe(List<SegmentToTranscribe> segmentsToTranscribe) {
     this.segmentsToTranscribe = segmentsToTranscribe;
   }
+
+   /**
+   * Returns the list of speaker names
+   *
+   * @return List of speaker names
+   */
+  public List<SpeakerName> getSpeakerNames() {
+    return speakerNames;
+  }
+
+  /**
+   * Specifies the list of speaker names in an audio file
+   * This property is optional but can only be used with "human" transcriber.
+   *
+   * @param speakerNames List of speaker names
+   */
+  public void setSpeakerNames(List<SpeakerName> speakerNames) {
+    this.speakerNames = speakerNames;
+  }
 }

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

@@ -0,0 +1,51 @@
+package ai.rev.speechtotext.models.asynchronous;
+
+import com.google.gson.annotations.SerializedName;
+
+/**
+ * A SpeakerInfo object represents the information related to the speaker in a transcript
+ * Used for speech-to-text jobs submitted to be transcribed by a human.
+ * */
+public class SpeakerInfo {
+  @SerializedName("id")
+  private String id;
+
+  @SerializedName("display_name")
+  private String displayName;
+
+  /**
+   * Returns the id of the speaker
+   *
+   * @return the id of the speaker
+   */
+  public String getIdName() {
+    return id;
+  }
+
+  /**
+   * Sets the id of the speaker
+   *
+   * @param displayName the displayed name of the speaker
+   */
+  public void setId(String id) {
+    this.id = id;
+  }
+
+  /**
+   * Returns the displayed name of the speaker
+   *
+   * @return the displayed name of the speaker
+   */
+  public String getDisplayName() {
+    return displayName;
+  }
+
+  /**
+   * Sets the displayed name of the speaker
+   *
+   * @param displayName the displayed name of the speaker
+   */
+  public void setDisplayName(String displayName) {
+    this.displayName = displayName;
+  }
+}

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

@@ -0,0 +1,30 @@
+package ai.rev.speechtotext.models.asynchronous;
+
+import com.google.gson.annotations.SerializedName;
+
+/**
+ * A object representing information provided about a speaker
+ * Used for speech-to-text jobs submitted to be transcribed by a human.
+ * */
+public class SpeakerName {
+  @SerializedName("display_name")
+  private String displayName;
+
+  /**
+   * Returns the displayed name of the speaker
+   *
+   * @return the displayed name of the speaker
+   */
+  public String getDisplayName() {
+    return displayName;
+  }
+
+  /**
+   * Sets the displayed name of the speaker
+   *
+   * @param displayName the displayed name of the speaker
+   */
+  public void setDisplayName(String displayName) {
+    this.displayName = displayName;
+  }
+}

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

@@ -7,6 +7,8 @@
 import ai.rev.speechtotext.models.asynchronous.RevAiJobOptions;
 import ai.rev.speechtotext.models.asynchronous.RevAiJobStatus;
 import ai.rev.speechtotext.models.asynchronous.RevAiJobType;
+import ai.rev.speechtotext.models.asynchronous.SpeakerName;
+import ai.rev.speechtotext.models.asynchronous.SegmentToTranscribe;
 import ai.rev.testutils.AssertHelper;
 import com.google.gson.Gson;
 import okhttp3.MediaType;
@@ -163,6 +165,20 @@ public void SubmitJobUrl_UrlAndOptionsAreSpecified_ReturnsARevAiJob() throws IOE
     options.setDeleteAfterSeconds(0);
     options.setLanguage("en");
     options.setTranscriber("machine_v2");
+    options.setVerbatim(true);
+    options.setRush(true);
+    options.setTestMode(true);
+    List<SegmentToTranscribe> segmentToTranscribeList = new ArrayList<>();
+    SegmentToTranscribe segment = new SegmentToTranscribe();
+    segment.setStartTimestamp(2.0);
+    segment.setEndTimestamp(100.5);
+    segmentToTranscribeList.add(segment);
+    options.setSegmentsToTranscribe(segmentToTranscribeList);
+    List<SpeakerName> speakerNamesList = new ArrayList<>();
+    SpeakerName speaker = new SpeakerName();
+    speaker.setDisplayName("Steve");
+    speakerNamesList.add(speaker);
+    options.setSpeakerNames(speakerNamesList);
 
     RevAiJob revAiJob = sut.submitJobUrl(options);
 
@@ -186,6 +202,20 @@ public void SubmitJobUrl_UrlAndOptionsAreSpecified_WithAuthHeaders_ReturnsARevAi
     options.setDeleteAfterSeconds(0);
     options.setLanguage("en");
     options.setTranscriber("machine_v2");
+    options.setVerbatim(true);
+    options.setRush(true);
+    options.setTestMode(true);
+    List<SegmentToTranscribe> segmentToTranscribeList = new ArrayList<>();
+    SegmentToTranscribe segment = new SegmentToTranscribe();
+    segment.setStartTimestamp(2.0);
+    segment.setEndTimestamp(100.5);
+    segmentToTranscribeList.add(segment);
+    options.setSegmentsToTranscribe(segmentToTranscribeList);
+    List<SpeakerName> speakerNamesList = new ArrayList<>();
+    SpeakerName speaker = new SpeakerName();
+    speaker.setDisplayName("Steve");
+    speakerNamesList.add(speaker);
+    options.setSpeakerNames(speakerNamesList);
 
     RevAiJob revAiJob = sut.submitJobUrl(options);