added AudioMerge for Speech Batch Transcript

GBB.ConversationalKM.Python/AudioMerge/__init__.py

@@ -0,0 +1,94 @@
+import logging
+import azure.functions as func
+from azure.storage.blob import BlobServiceClient, BlobClient, ContainerClient
+import json
+import os
+from isodate import parse_duration
+import dateutil.parser
+
+
+def main(myblob: func.InputStream):
+    logging.info(f"Python blob trigger function processed blob \n"
+                 f"Name: {myblob.name}\n"
+                 f"Blob Size: {myblob.length} bytes")
+
+    connection_string = os.environ['conversationalkm_STORAGE']
+    container = os.environ['telemetry_processed']
+
+    blob_service_client = BlobServiceClient.from_connection_string(connection_string)
+
+    # Processing
+    get_transcript(myblob, blob_service_client, container)
+
+
+def get_transcript(blob, blob_service_client, container):
+    # speechtotext/v3.0/transcriptions/{id}
+
+    transcript_id = blob.name.split('/')[-1]
+
+    transcript= json.loads(blob.read())
+
+    timestamp = transcript['timeStamp']
+    recognizedPhrases = transcript['recognizedPhrases']
+    phrases = list(map(lambda x: extract_data(x, timestamp), recognizedPhrases))
+
+    save_conversation(transcript_id, phrases, blob_service_client, container)
+
+# Extract date from Speech Data
+def extract_data(x, timestamp):
+    return {
+        "speaker" : x['channel'],
+        "phrase" : x['nBest'][0]['display'],
+        "offset" : x['offset'],
+        "duration": x['duration'],
+        "offsetInTicks" : x['offsetInTicks'],
+        "durationInTicks" : x['durationInTicks'],
+        "timestamp" : timestamp
+    }
+
+
+# Format single phrase as message
+def format_message(id, phrase):
+    event_time = dateutil.parser.parse(phrase['timestamp']) + parse_duration(phrase['offset'])
+    return  {
+        "Id": f"{id}_{phrase['offset']}",
+        "ReferenceId": None,
+        "EventType": "MessageFromUser" if phrase['speaker'] == 0 else "MessageFromBotOrAgent",
+        "EventTime": str(event_time),
+        "ConversationId": id,
+        "Value": phrase['phrase'],
+        "UserId": phrase['speaker'],
+        "CustomProperties": {
+            "offset": phrase['offset'],
+            "duration": phrase['duration'],
+            "offsetInTicks": phrase['offsetInTicks'],
+            "durationInTicks": phrase['durationInTicks']
+            }
+        }
+
+
+# Save conversation in Azure Blob Storage
+def save_conversation(conversation_id, messages, blob_service_client, container):
+    # Define structure for conversational data
+    data = {
+        "ConversationId" : conversation_id,
+        "Messages" : []
+    }
+    blob = f"{conversation_id}.json"
+    # Define Azure Blob Storage client
+    blob_client = blob_service_client.get_blob_client(container=container, blob=blob)
+
+    data['Messages'] = list(map(lambda x: format_message(conversation_id, x),messages))
+    data['Messages'] = sorted(data['Messages'], key = lambda x:x['EventTime'])
+
+
+    # Extract Min and Max EventTime for conversation
+    data["StartTime"] = min(list(map(lambda x: x['EventTime'], data['Messages'])))
+    data["EndTime"]   = max(list(map(lambda x: x['EventTime'], data['Messages'])))
+
+    data["merged_content"]         = "".join(list(filter(None, map(lambda x: x['Value'] , data['Messages']))))
+    data["merged_content_user"]    = "".join(list(filter(None, map(lambda x: x['Value'] if x['UserId'] == 0 else None, data['Messages']))))
+    data["merged_content_agent"]   = "".join(list(filter(None, map(lambda x: x['Value'] if x['UserId'] == 1 else None, data['Messages']))))
+
+    # Write results on Azure Blob Storage
+    blob_client.upload_blob(json.dumps(data), overwrite=True)

GBB.ConversationalKM.Python/AudioMerge/function.json

@@ -0,0 +1,12 @@
+{
+    "scriptFile": "__init__.py",
+    "bindings": [
+      {
+        "name": "myblob",
+        "type": "blobTrigger",
+        "direction": "in",
+        "path": "json-result-output/{name}",
+        "connection": "conversationalkm_STORAGE"
+      }
+    ]
+  }