Re: (Long post) Metaphone Algorithm In AWK

porkchop@invalid.foo (Mike Sanders) writes:

Hi folks, hope you all are doing well.
>
Please excuse long post, wanted to share this, some might find
it handy given a certain context. Must run, I'm very behind in
my work (hey I'm always running behind!)

Using a word list, I found some odd matches.  For example:

$ echo "drunkeness indigestion" | awk -f metaphone.awk -v find=texas
drunkeness
indigestion

Are these really metaphone matches for "texas"?  It's possible (I don't
know the algorithm at all well) but I found it surprising.

# metaphone.awk: Michael Sanders - 2024
#
# example invocation:
#
# echo "texas taxes taxi" | awk -f metaphone.awk -v find=texas
#
# notes:
#
# ever notice when you search for (say):
#
# 'i went to the zu'
#
# & your chosen search engine suggests something like:
#
# 'did you mean i went to the zoo'
#
# the metaphone algorithm handles such cases pretty well actually...
#
# Metaphone  is a phonetic algorithm, published by Lawrence Philips in
# 1990,   for  indexing  words  by  their  English  pronunciation.  It
# fundamentally improves on the Soundex algorithm by using information
# about   variations  and  inconsistencies  in  English  spelling  and
# pronunciation  to  produce  a  more  accurate encoding, which does a
# better job of matching words and names which sound similar.
# https://en.wikipedia.org/wiki/Metaphone
#
# english only (sorry)
#
# not extensively tested, nevertheless a solid start, if you
# improve this code please share your results
#
# other implentations...
#
# gist:  https://gist.github.com/Rostepher/b688f709587ac145a0b3
#
# BASIC: http://aspell.net/metaphone/metaphone.basic
#
# C:     http://aspell.net/metaphone/metaphone-kuhn.txt

I wanted a "reference" implementation I could try, but this is not a
useful C program.  It's in a odd dialect (it uses void but has K&R
function definitions) and has loads of undefined behaviours (strcpy of
overlapping strings, use if uninitialised variables etc).

# check if a character is a vowel
function isvowel(c, is_vowel) {
  is_vowel = c ~ /[AEIOU]/
  return is_vowel
}

I was not going to comment on the code, but this hit me just before I
posted.  Given the odd way AWK functions have to define locals, I tend
to use them only when really needed.  Here I think I would just write

function isvowel(c) {
   return c ~ /[AEIOU]/
}

--
Ben.