[net.sources] Spelling improvement tools and ideas

bobvan (11/10/82)

A recent net.general article asked if there wasn't something that could
be done to improve the poor spelling seen on the net.  The code presented
here won't cure poor spelling, but can help those who want to improve.

The first thing you should do is to get in the habit of using spell(1)
regularly.  A few lines in your bin directory can make the benefit even
greater.  Simply create your own version of spell that logs all errors
detected.  This should look something like:

	#! /bin/sh
	spell $* | tee -a $HOME/spell.errors

Users of vispell need something like this:

	#! /bin/sh
	tee /tmp/vis$$
	echo SpellingList
	spell /tmp/vis$$ | tee -a $HOME/spell.errors
	rm /tmp/vis$$

Then you should run the following pipeline about once a month to
get a list of your most common spelling mistakes.

	sort -d $HOME/spell.errors | uniq -c | awk '$1 > 1' | sort -rn

System administrators can assist by taking advantage of the log kept
by spell.  The file /usr/dict/spellhist contains a list of all words
rejected by the spell.  Once per month, I run a pipeline like the one
above on this file.  I add any legitimate words or local abbreviations
to the dictionary.  This reduces the number of "false hits" from spell,
thereby encouraging people to use it more regularly.  I generaly put
the most commonly misspelled words into /etc/motd for a few days.
I am no expert on spelling or word usage, so I ask for help when making
these judgements.

Here is the pipeline that I run with the monthly accounting:

	: Spelling history
	:
	egrep -v "tty??|tty[0-9][0-9]|ttyd[0-9]" /usr/dict/spellhist |\
		sort -d |\
		uniq -c |\
		awk '$1 > 4' |\
		sort -rn |\
		tee -a /usr/dict/cumhist |\
		pr -h 'MONTHLY:  Most frequently misspelled words' |\
		lpr
	cp /dev/null /usr/dict/spellhist

Interestingly enough, the most frequently misspelled "word" list last month
was "SpellingList" -- we are heavy vispell users at this site.

				Bob Van Valzah
				(...!decvax!ittvax!tpdcvax!bobvan)