Call transcription

Call recording is the capture and storage of audio from a customer call. Companies keep recordings for quality monitoring, agent training, dispute resolution, and regulatory compliance. It differs from call transcription, which converts that audio into text. Recordings preserve tone and exact wording. Transcripts make conversations searchable. US consent law varies by state. Some states require only one party’s consent to record. Others require all parties to agree. Retention periods and encryption at rest should match company policy and applicable regulation. Payment calls should pause recording during card capture to keep cardholder data out of storage. When a customer disputes what was agreed on a cancellation call, the stored recording settles the facts. OnClarity’s AI Agent QA scores recorded conversations against existing evaluation rubrics and produces audit-trail exports for compliance review.

Word error rate (WER) measures speech recognition accuracy. It is the sum of substitutions, insertions, and deletions, divided by the total words in the reference transcript, expressed as a percentage. Lower is better. Vendor-published WER figures are only meaningful when compared on identical audio. Accents, background noise, telephony codecs, and domain-specific vocabulary all push the error rate higher in production, regardless of benchmark claims. A contact center example: a voicebot transcribes a caller’s accent clearly but consistently mishears a product or brand name. This inflates substitution errors until that term is added to a custom vocabulary list. This voice AI glossary will keep growing as new contact center infrastructure terms and voice AI terminology enter vendor conversations. Bookmark it alongside OnClarity’s main glossary of call center technology definitions and AI contact center glossary entries for the next RFP or vendor call.

Speech analytics automates analysis of recorded or live customer calls. It uses speech-to-text transcription combined with natural language processing to extract keywords, topics, and sentiment. Many systems also measure acoustic signals such as silence, talk-over, and speaking pace, to flag stress or disengagement that word choice alone would miss. Real-time speech analytics prompts agents mid-call and raises live compliance flags. Post-call analysis surfaces trends and root causes across a larger call sample. Standard uses include detecting required compliance phrases, categorizing contact reasons, and feeding evidence into quality assurance scoring.

PII redaction is the automatic removal of personally identifiable information from transcripts, logs, and training samples. It lets support conversations be analysed, stored, and used to improve models without carrying customer identifiers into systems that were never scoped to hold them.