How Does Anybody Type That Fast? A Keyboard With Twenty Two Keys

Speech runs at roughly one hundred and fifty to two hundred words per minute in ordinary conversation and considerably faster in argument, while a very good touch typist manages around eighty to a hundred and cannot sustain that indefinitely. The gap is unbridgeable by typing faster, so the solution changes what is typed. A machine with a small number of keys is operated by pressing several at once, and each such combination writes a whole syllable rather than one letter, which reduces the number of movements by a factor of several. Certified practitioners work at two hundred and twenty five words per minute and above, sustained, with accuracy requirements above ninety five per cent.

The layout is unlike an ordinary keyboard in every respect:

• •Around twenty two keys rather than a hundred
• •Consonants on the left for the start of a syllable
• •Vowels on a bar operated by the thumbs
• •Consonants on the right for the end of a syllable
• •Keys pressed simultaneously as a chord, producing one output
• •Output written phonetically rather than in standard spelling

The record produced at the machine is not readable by anybody untrained, which creates a second stage of work. Historically the practitioner read back their own phonetic notes and dictated or typed a full transcript afterwards, which took considerably longer than the proceedings themselves. Software now translates the chords into words in real time, using a dictionary that the individual practitioner builds and maintains, since each person develops their own abbreviations for names, terms and phrases they encounter often. That personal dictionary is what makes the output accurate, and preparing it before a trial by entering the names and technical vocabulary involved is a substantial part of the preparation.

The same technique supports several jobs that have nothing to do with courts. Live captioning for deaf and hard of hearing audiences at lectures, conferences, theatres and on television is produced by exactly these practitioners working in real time, which is the main growth area of the profession. Parliamentary and committee reporting uses it. Medical and academic transcription draws on it. Broadcast subtitling for live programmes uses either a practitioner on the machine or a technique in which somebody repeats what is said into a speech recognition system trained on their own voice, which is faster to learn and less accurate. Automatic speech recognition has improved enormously and continues to fail on crosstalk, accents and specialist vocabulary.

Courts in many jurisdictions have moved to audio recording with later transcription, which is cheaper, and the arguments about that are practical rather than sentimental. A person present can ask a speaker to repeat, can indicate who is talking when several speak at once, and can produce an immediate readback that a judge asks for, none of which a microphone does. Recordings capture room noise, miss speakers who turn away and fail without anybody noticing. Against that, recording costs far less and machines do not tire. The practical outcome in most places is a mixture, with recording used routinely and a person present for the most serious proceedings, and the profession is shrinking while retaining a role in live captioning for deaf audiences.

Speech outruns ordinary typing, so the machine writes syllables rather than letters, with around twenty two keys pressed in chords and output written phonetically. Practitioners sustain above two hundred and twenty five words per minute. Software translates the chords using a dictionary each person builds individually, and preparing it with a case's names and terms is substantial work. Audio recording has replaced the role in many routine proceedings.