[TUHS] Bell Labs sed performance

[Douglas McIlroy via TUHS]

 > I put together a small utility that extracts “unusual” words from a text

This reminds me of a Lorinda Cherry story.

Lorinda made a suite of programs to search records of AT&T
customer-service calls for systemic anomalies evidenced by sudden
changes in word frequency.

 One amusing catch was "ten". This mundane word suddenly spiked in
Missouri. It seems that the automated instructions for prepaid
phone-card calls had been updated to say something like "Please enter
your card number and PIN". In Missouri, where many people pronounce
"ten"  as "tin", this led to folks complaining that their calls did
not go through although they were sure they had correctly entered the
card number and "tin".

Lorinda's biggest triumph was "monkey". A newsletter that AT&T
distributed to some customers carried a piece about international
calling, illustrated by a sketch map of the world that showed a
cartoon person on each continent with connections between them.
Offended customers called to complain that the person in Africa looked
like a monkey. This complaint  rocked the executive suite. Justly
chastened, AT&T abolished the newsletter.

Doug