問題文
A call analytics solution must produce searchable transcripts with sensitive values removed. In what order should the steps run?
選択肢
- Redact the audio before transcription by muting the segments where a speaker is likely to be reading out a number, then transcribe the muted recording
- Translate the transcript into a common language and then index it, so that calls in different languages can be searched with a single set of query terms
- Transcribe the audio, detect and redact the sensitive values, and only then index the redacted version of the text rather than the original wording of it
- Index the transcript first so the calls become searchable as early as possible, then detect the sensitive values and redact them inside the index in a follow-up pass