Роутинг звонков, отсеивание отказов и метрики
Роутинг звонков, отсеивание отказов и метрики
Ответить самому
Сначала сформулируйте ответ как на собеседовании, затем откройте разбор и оцените себя.
Короткий ответ
Use a cheap accept/reject detector before expensive extraction, track precision/recall for accepted calls, branch accuracy, slot accuracy and end-to-end booking correctness.
Полный разбор
A high-volume pipeline should not send every call to the most expensive model. First run cheap filters: duration thresholds, VAD, ASR with small model if possible, and a lightweight accepted/rejected classifier. Rejected or very short calls can bypass branch/time extraction unless confidence is low.
Metrics should match business damage. For accepted/rejected, use precision and recall with a clear preference: missing a real booking loses revenue, while booking a rejected customer creates an operational error. For extracted fields, track exact branch accuracy, exact slot accuracy, date accuracy and full booking accuracy.
Also track coverage. If the system returns “needs review” too often, it may be accurate but not useful. Segment metrics by operator, branch, language, call duration and ASR quality because failures will cluster.
Теория
A routing layer makes expensive extraction scalable and lets each stage use metrics appropriate to its decision.
Типичные ошибки
- Run the LLM on all calls without cheap rejection filtering.
- Report one aggregate accuracy for different failure modes.
- Ignore abstention or human-review rate.
- Forget that false bookings and missed bookings have different costs.
Как отвечать на собеседовании
- Separate accept/reject metrics from field extraction metrics.
- Mention early exits for short rejections.