Better PatternCount?

[mweiss]

My db consists of transcripts of interviews. For analytical purposes, I want to count how frequently certain words appear. The PatternCount function almost, but not quite, gets me what I want. The problem is this: Suppose I want to know how often the word "my" appears in a line of transcript. I can use

PatternCount(Transcript::Text ; "my")

But this results in a lot of "false positives", e.g. "academy" and "Jeremy". I could eliminate these by using the search string " my" (with a prepended space), but then I miss any time the word "my" appears at the beginning of the text.

Surely there must be some way to do a pattern count for whole words, right? Or some well-known workaround?

[Fitch]

Add a space before the text to search and that takes care of at least the problem you mentioned:

PatternCount(" " & Transcript:: Text ; " my")

You could also add a space to the end the same way. In addition you could enclose the first parameter in a Substitute function to turn all carriage returns, periods, commas etc. into spaces:

PatternCount(

Substitute( " " & Transcript:: Text & " " ; [",";" "] ; [".";" "] ; [etc.] ) ;

" my ")

If you're going to reuse the calc, you might consider making it a custom function.