FREQ

[Brian MacWhinney]

Dear ChiBolts,

    I am replying here to a message that Darinka posted to info-childes.  Because this is a message about the functioning of CLAN programs, it is best discussed on chibolts instead.  Here is her message and my answer:

Dear collegues,

 

I wanted to match codes on the %mor tier to words on the main line and I used MODREP for it:

 

MODREP +u +k +r2 +b%mor +c* +t@ID=2.*.*.*.*.*  +o"v|*" *.mor

 

But then I found out that the output of this command do not provide the same frequency for codes as I get from FREQ on %mor tier. MODREP provided systematically larger frequency then FREQ.

 

I had expected them to be exactly the same. Do I missunderstand something?

 

 

Darinka Andjelkovic

 

Laboratory of Experimental Psychology and Institute of Psychology

Faculty of Philosophy

University of Belgrade

Serbia

Dear Darinka,

     The difference arises from the fact that the %mor line excludes repetitions and several other things from the main line.  The goal of the %mor line is to analyze morphology and syntax.  To do this, excluding repetitions helps in automatic part of speech tagging.  The %mod line, on the other hand, is pegged to the main line in a way that includes all words, including repetitions.  Therefore, it will have larger numbers than a FREQ on the %mor line.  You can modify what FREQ looks at on the main line by using the +r6 switch to exclude repetitions.  You might check that.  In general, few people have used the %mod line as the target for FREQ, so we haven't payed much attention to the details of what is there.  

-- Brian MacWhinney

[Leonid Spektor]

Darinka,

If you want to get frequencies count of codes on %mor tier and their corresponding words from speaker tier, then you can use FREQ command with +d7 option. You will need to get CLAN that is less that about 9 month old. The command line is:

freq +s@|-v,r-*,-+*,&+*,#+*,=+*,o-% +d7 *.cha

Leonid.