Hi Folks,  I have been watching this thread for awhile.  The subject has
interested me for a long time and I still approach it with some
enthusiasm, although much more dampened than in earlier times.

My experience is that this project is far more difficult than it first
appears.

The previously-mentioned three books (The Up/Down book and the two by
Barlow & Morganstern) have been of some value to me, but results have
been far from ideal for a number of reasons:

1. Popular hits of the time in which we are interested turn up
infrequently in the above books.

2. Tunes are often heavily arranged and it can be difficult to separate
ornamentation from theme.  With an alphabetized or up & down listing you
can be thrown off as soon as the first or second note.

3. Tunes sometimes don't start at the "official" beginning and verse and
chorus may not be obviously separated.

More than ten years ago, while still believing that all things were
possible if I tried hard enough (a philosophy of mine which has been
changing over time), I started a computerized database using the up/down
model.  With a great deal of help from my friend Olin Tillotson we even
presented a couple of workshops on the subject at MBSI conventions.

Although the project has been on a back burner for a number of years, by
far the best source of identification is still Olin with his fine musical
ear, memory and ability to recognize a tune as one of the unidentified
when it turns up on the radio.

I will be most interested to follow this thread to see how others are
making out with it.  I also want to offer a warning that what seems to be
a relatively straight-forward task is anything but, and that anyone
seriously taking it on should be prepared to spend a lot of time.

This brings us to the related thread being discussed, which is the
transcription of musical box cylinder pin placement to a MIDI recording
or other digital representation.

 [ I split this topic to a separate article. -- Robbie ]

In terms of tune identification, all of the above problems will, of
course, still exist and I doubt that simple computer programming will be
able to get around most of them.

Jim Heyworth,
Sechelt, B.C., Canada

 [ I think it was Larry Smith, several months ago, who described how
 [ new computer algorithms such as Fuzzy Logic might be used for "best
 [ guess" pattern matching.  In any event, the ultimate machine would
 [ be the hit of a cocktail party: just sing the unknown song into the
 [ microphone for "The Trivial Music Game"  ;-)
 [ -- Robbie