Answer Seizure Rate (ASR) vs Automatic Speech Recognition (ASR) — Answer Seizure Rate / Automatic Speech Recognition

A seizure is an attempt: the moment the originating switch grips a trunk and sends the call toward the destination. Answer Seizure Rate divides the calls that were answered by those seizures, usually as a percentage, and carriers and wholesale operators read it per route exactly as they read PDD, jitter, packet loss, and MOS for the media path. ITU-T E-series traffic engineering treats ASR, ACD (Average Call Duration), and PDD as the route-health triad; a route whose ASR slips while its audio stays clean is still a route to re-route around.

Automatic Speech Recognition is the unrelated, same-acronym term from machine learning: the speech-to-text step at the front of a voice-agent pipeline, paired with text-to-speech at the other end. Because the collision lands in real product surfaces, this entry is the disambiguation page: the STT (Speech-to-Text) entry (see S in the docs glossary) defines the transcription technology, while the telephony metric pairs with the disconnect-reason ladder (busy, declined, no answer, unrouted) that moves the dial-side number.

In Orbit, the machine-learning meaning is the one inside the product: the speech-recognition spend line on the LLM-spend budget and the ASR entity correction on call transcripts both name Automatic Speech Recognition, and the four STT providers in an agent's voice settings failover chain (Deepgram Nova-3, OpenAI Whisper, Azure Speech, Google STT) are the same technology. Orbit's telephony voice-quality surfaces document the rest of the triad, PDD, MOS (Mean Opinion Score), jitter, and packet loss, and calls ride the wholesale softswitch described in the jambonz-softswitch concept. Orbit does not compute or display an Answer Seizure Rate KPI today; voice operators reading the term on a carrier report or a cost line should read it in the telephony sense.