Question

A telecommunications company runs a call center that records thousands of customer support calls daily in various languages. The architecture team wants to build an end-to-end pipeline that transcribes each recorded call, detects the sentiment of the conversation, and stores structured results for downstream analytics. The team has no data scientists and wants to avoid training or hosting custom models. Which approach using the Gemini Enterprise Agent Platform and Google Cloud prebuilt APIs best meets these requirements?

Accepted Answer

Use the prebuilt Speech-to-Text API to transcribe audio and the Natural Language API to analyze sentiment, orchestrating the flow with Cloud Functions triggered by uploads to Cloud Storage.

Answer

Train a custom AutoML audio classification model on Vertex AI, deploy it to an endpoint, and invoke it from a scheduled Dataflow job.

Answer

Build a custom TensorFlow model for transcription and sentiment, train it on GPUs in Compute Engine, and serve predictions behind a load balancer.

Answer

Use the Video Intelligence API to process the audio recordings and export the detected transcript and sentiment to BigQuery.