Added Transcription Packet Parser.

sdk/communication/communication-call-automation/src/models/transcription.ts

@@ -0,0 +1,70 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT license.
+
+import { CommunicationIdentifier } from "@azure/communication-common";
+
+/**
+ * The status of the result of transcription.
+ */
+export enum ResultStatus {
+  /** Intermediate result.*/
+  Intermediate = "intermediate",
+  /** Final result.*/
+  Final = "final",
+}
+
+/**
+ * The format of transcription text.
+ */
+export enum TextFormat {
+  /** Formatted recognize text with punctuations.*/
+  Disply = "display",
+}
+
+/**
+ * Text in the phrase.
+ */
+export interface WordData {
+  /** Text in the phrase.*/
+  text: string;
+  /** The word's position within the phrase.*/
+  offset: number;
+  /** Duration in ticks. 1 tick = 100 nanoseconds.*/
+  duration: number;
+}
+
+/**
+ * Metadata for Transcription Streaming.
+ */
+export interface TranscriptionMetadata {
+  /** Transcription Subscription Id.*/
+  subscriptionId: string;
+  /** The target locale in which the translated text needs to be.*/
+  locale: string;
+  /** call connection Id.*/
+  callConnectionId: string;
+  /** correlation Id.*/
+  correlationId: string;
+}
+
+/**
+ * Streaming Transcription.
+ */
+export interface TranscriptionData {
+  /** The display form of the recognized word.*/
+  text: string;
+  /** The format of text.*/
+  format: TextFormat;
+  /** Confidence of recognition of the whole phrase, from 0.0 (no confidence) to 1.0 (full confidence). */
+  confidence: number;
+  /** The position of this payload. */
+  offset: number;
+  /** Duration in ticks. 1 tick = 100 nanoseconds. */
+  duration: number;
+  /** The result for each word of the phrase. */
+  words: WordData[];
+  /** The identified speaker based on participant raw ID. */
+  participant: CommunicationIdentifier;
+  /** Status of the result of transcription. */
+  resultStatus: ResultStatus;
+}

sdk/communication/communication-call-automation/src/utli/streamingDataParser.ts

@@ -0,0 +1,43 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT license.
+
+import { createIdentifierFromRawId } from "@azure/communication-common";
+import { TranscriptionMetadata, TranscriptionData } from "../models/transcription";
+
+/** Parse the incoming package. */
+export function streamingData(
+  packetData: string | ArrayBuffer,
+): TranscriptionMetadata | TranscriptionData {
+  let stringJson: string;
+  if (typeof packetData === "string") {
+    stringJson = packetData;
+  } else {
+    const decoder = new TextDecoder();
+    stringJson = decoder.decode(packetData);
+  }
+
+  const jsonObject = JSON.parse(stringJson);
+  const kind: string = jsonObject.kind;
+
+  switch (kind) {
+    case "TranscriptionMetadata": {
+      const transcriptionMetadata: TranscriptionMetadata = jsonObject.transcriptionMetadata;
+      return transcriptionMetadata;
+    }
+    case "TranscriptionData": {
+      const transcriptionData: TranscriptionData = {
+        text: jsonObject.transcriptionData.text,
+        format: jsonObject.transcriptionData.format,
+        confidence: jsonObject.transcriptionData.confidence,
+        offset: jsonObject.transcriptionData.offset,
+        duration: jsonObject.transcriptionData.duration,
+        words: jsonObject.transcriptionData.words,
+        participant: createIdentifierFromRawId(jsonObject.transcriptionData.participantRawID),
+        resultStatus: jsonObject.transcriptionData.resultStatus,
+      };
+      return transcriptionData;
+    }
+    default:
+      throw new Error(stringJson);
+  }
+}

sdk/communication/communication-call-automation/test/streamingDataParser.spec.ts

@@ -0,0 +1,69 @@
+// Copyright (c) Microsoft Corporation.
+// Licensed under the MIT license.
+
+import { TranscriptionData, TranscriptionMetadata } from "../src/models/transcription";
+import { streamingData } from "../src/utli/streamingDataParser";
+import { assert } from "chai";
+
+describe("Stream data parser unit tests", function () {
+  const encoder = new TextEncoder();
+  const transcriptionMetaDataJson =
+    '{"kind":"TranscriptionMetadata","transcriptionMetadata":{"subscriptionId":"0000a000-9999-5555-ae00-cd00e0bc0000","locale":"en-US","callConnectionId":"6d09449c-6677-4f91-8cb7-012c338e6ec1","correlationId":"6d09449c-6677-4f91-8cb7-012c338e6ec1"}}';
+  const transcriptionDataJson =
+    '{"kind":"TranscriptionData","transcriptionData":{"text":"Hello everyone.","format":"display","confidence":0.8249790668487549,"offset":2516933652456984600,"words":[{"text":"hello","offset":2516933652456984600},{"text":"everyone","offset":2516933652459784700}],"participantRawID":"4:+910000000000","resultStatus":"Final"}}';
+
+  it("Successfully parse binary data to transcription meta data ", function () {
+    const transcriptionMetaDataBinary = encoder.encode(transcriptionMetaDataJson);
+    const parsedData = streamingData(transcriptionMetaDataBinary);
+    if ("locale" in parsedData) {
+      validateTranscriptionMetadata(parsedData);
+    }
+  });
+
+  it("Successfully parse json data to transcription meta data ", function () {
+    const parsedData = streamingData(transcriptionMetaDataJson);
+    if ("locale" in parsedData) {
+      validateTranscriptionMetadata(parsedData);
+    }
+  });
+
+  it("Successfully parse binary data to transcription data ", function () {
+    const transcriptionDataBinary = encoder.encode(transcriptionDataJson);
+    const parsedData = streamingData(transcriptionDataBinary);
+    if ("text" in parsedData) {
+      validateTranscriptionData(parsedData);
+    }
+  });
+
+  it("Successfully parse json data to transcription data ", function () {
+    const parsedData = streamingData(transcriptionDataJson);
+    if ("text" in parsedData) {
+      validateTranscriptionData(parsedData);
+    }
+  });
+});
+
+function validateTranscriptionMetadata(transcriptionMetadata: TranscriptionMetadata): void {
+  assert.equal(transcriptionMetadata.subscriptionId, "0000a000-9999-5555-ae00-cd00e0bc0000");
+  assert.equal(transcriptionMetadata.locale, "en-US");
+  assert.equal(transcriptionMetadata.correlationId, "6d09449c-6677-4f91-8cb7-012c338e6ec1");
+  assert.equal(transcriptionMetadata.callConnectionId, "6d09449c-6677-4f91-8cb7-012c338e6ec1");
+}
+
+function validateTranscriptionData(transcriptionData: TranscriptionData): void {
+  assert.equal(transcriptionData.text, "Hello everyone.");
+  assert.equal(transcriptionData.resultStatus, "Final");
+  assert.equal(transcriptionData.confidence, 0.8249790668487549);
+  assert.equal(transcriptionData.offset, 2516933652456984600);
+  assert.equal(transcriptionData.words.length, 2);
+  assert.equal(transcriptionData.words[0].text, "hello");
+  assert.equal(transcriptionData.words[0].offset, 2516933652456984600);
+  assert.equal(transcriptionData.words[1].text, "everyone");
+  assert.equal(transcriptionData.words[1].offset, 2516933652459784700);
+  if ("kind" in transcriptionData.participant) {
+    assert.equal(transcriptionData.participant.kind, "phoneNumber");
+  }
+  if ("phoneNumber" in transcriptionData.participant) {
+    assert.equal(transcriptionData.participant.phoneNumber, "+910000000000");
+  }
+}