A 311 call center wants residents to dictate service requests over a phone IVR that feeds an agentic intake flow. Which speech capability should they implement first?
Select an answer to reveal the explanation.
Short Explanation
Phone dictation is like a stenographer on the line with the resident. Speech-to-text turns spoken 311 requests into text the agentic intake flow can actually use. Text-to-speech confirmations alone, a translation step with no transcript, or speaker-ID verification never hears the caller.
Full Explanation
The 311 call center wants residents to dictate service requests over a phone IVR that feeds an agentic intake flow. The first speech capability to implement is speech-to-text so spoken requests become text the agentic intake flow can parse, classify, and route. Without transcription, the agent has no textual representation of the resident’s pothole, trash, or noise complaint to act on.
Text-to-speech to speak case confirmations back to residents during the IVR call flow is outbound audio only; it confirms cases after intake rather than capturing resident speech or creating intake text. Speech translation that converts the resident's words into another language before intake still requires recognizing the spoken audio first, and translating it does not by itself hand text to the agentic intake flow in the caller's own words. Speaker recognition to verify resident identity by voice before processing the request keeps the phone path but authenticates the caller instead of transcribing the service request into text the agentic intake flow can use. Each distractor either speaks without capturing the request, verifies identity without transcribing it, or converts speech into another language without handing text to the intake flow in the caller’s own words.
Exam caveat: for phone intake into agents, STT is the ingress modality; TTS may still be used for prompts and confirmations, but the stem’s “dictate … feeds an agentic intake flow” points to STT first. Do not pick speech translation unless multilingual conversion is required.
Operational check: place test calls covering accents and street names common to the city, measure STT word error rate on addresses and issue types, confirm transcribed text lands in the agent intake schema, and require confirmation read-back before creating a work order. Log confidence scores and offer keypad fallback when recognition is poor.
Prefer to study offline? Printable PDF study book + test engine ($7.50, unofficial). These questions stay free here.