Realtime Lag, a Dying Recorder, and a Soft-Spoken Advocate
You are the stenographic reporter and realtime feed provider for a three-hour arbitration. Two remote participants receive your realtime: one on a law-firm LAN, one on a 4G hotspot. Twenty minutes into the morning session the hotspot user messages that text is arriving in bursts, roughly 40 seconds behind. You are writing about 220 words per minute, spiking to 260 during cross-examination. Your writer saves to a USB stick; the backup audio recorder is a separate device showing 20 percent battery, and there is no spare. One advocate speaks softly with a heavy accent and is the one doing most of the examining. A ten-minute recess is scheduled in twenty minutes. Your CAT software's realtime send buffer is set high.
1. Rank your actions over the next twenty minutes before the recess and justify the order.
Read the worked answer
1. Communicate first. Tell the hotspot user plainly that the feed is degraded, that text will be late, and that they should not rely on it for anything time-sensitive. An unannounced lag is worse than a known one, because the consumer may act on stale text.
2. Attack latency at the source, not accuracy. Reduce the send buffer and enable the flush/send-now setting so completed text leaves immediately rather than waiting for a full buffer. Do not drop the buffer to zero: you will fragment sentences and push out half-translated strokes, which trades one defect for a worse one.
3. Give the affected consumer a fallback. Offer an audio bridge or dial-in so they have a live channel while the data path recovers. Realtime is a convenience; the record does not depend on it.
4. Protect the primary record. Do not restart the writer, re-seat the USB stick, or experiment mid-session. Replace the recorder's battery at the recess if any spare exists anywhere in the building; otherwise shorten the risk window by starting a second recording device (phone voice memo) as a stopgap and note the time it began.
Ordering rationale: notification and fallback are reversible and cost nothing; buffer changes are reversible; hardware changes are not, so they wait for the recess.
2. At the recess you find the recorder's file is unreadable from minute 74 to 81, covering the soft-spoken advocate's key exchange. Your steno notes for that span are intact and legible. What is the record, what do you produce, and what do you disclose?
Read the worked answer
The notes are the record. When you are the stenographic officer, the audio file is your working backup, not the official record. A corrupt eight-minute span in a backup file creates no gap in the proceedings.
- Produce from your notes, translated normally, with no ellipses or [inaudible] insertions merely because the audio failed. Inserting them would misrepresent the record as incomplete when it is not.
- Document internally: note the file name, the time span, and that the notes are the source, in your job file. If you later re-transcribe for any reason, the discrepancy is explained.
- Disclose selectively and neutrally. If any passage in that span is genuinely uncertain to you - a proper name, a number, a term of art - raise it through the ordinary read-back or correction procedure rather than volunteering a general announcement that 'the audio failed.' A blanket disclosure invites a stipulation that the record is incomplete when it is not.
- Never guess and never silently paper over. If you cannot resolve a term, mark it and say so; a flagged uncertainty is defensible, an invented word is not.
3. How do you handle the accented, soft-spoken examining advocate while writing live?
Read the worked answer
Separate two problems that feel like one: volume and accent.
- Volume is fixable and urgent - ask for a microphone, ask counsel to speak up, or ask to move the recorder and yourself closer. Do it early and once, on the record if necessary, so the request is not mistaken for partisanship later.
- Accent is not a defect. The task is to render what a fluent listener would hear, in standard orthography, not to phonetically imitate or simplify. Do not drop articles, do not write 'dis' for 'this,' and do not flatten syntax because the speaker's first language is not English. That is a common and serious error, and it is far more damaging than a [inaudible].
- Build a name and term list in real time. Most mishearing of accented speech is mishearing of proper nouns and technical terms, not of ordinary words. Capture spellings at the first break.
- When you truly lose a phrase, mark it and resolve it at the next recess by read-back to the speaker, not by reconstruction from context.