The following list outlines the various scripts and programs developed by members of the LDP.
A simple script that updates all svn repositories that are being used. This was a lot more useful before the repositories were consolidated, but this can still be run from anywhere.
A simple script that opens the specified .cut file from the $CLAN/lib/english/lex in a vi editor.
Identifies all capitalized words and all ampersanded words in a CHAT file. Also identifies those capitalized words that also appear in lower case and those ampersanded words that also appear without an ampersand.
Usage:
caps xx.xx.cha
Sample output:
Capitalized words:
Alls
Bam --> check lower case
Caribou
Dad
Daddy
Flipper
Fuh
Ick --> check lower case
Kuh --> check lower case
La
Lilia
Mama --> check lower case
Mexican
Mom
Mommy
Museum_Of_Science_And_Industry
Pete
Seesee
Uhuh
Vroom --> check lower case
Vuh --> check lower case
Wah --> check lower case
Zs
Ampersanded words:
&fa --> check for unampersanded instances
&fuh
&la
A simple script that changes directory to the $CLAN/lib/english/lex directory, commits any changes to the svn repository, and returns to the previous working directory.
Compares the tags of specified tiers in chat files.
The script compares a test tier against a gold standard tier. The tiers to be compared can be in the same file or separate files.
To compare two syntax tiers in the same file (‘test.cha’):
compare.py --test TIER --gold TIER test.cha
If comparing seperate files (‘test.cha’ against ‘gold.cha’):
compare.py --test TIER --gold TIER test.cha gold.cha
Note that both –test and –gold default to ‘syn’:
compare.py test.cha gold.cha
-h, --help show this help message and exit
-g <TIER>, --gold=<TIER> tier to use as gold standard
-t <TIER>, --test=<TIER> tier to test against gold standard
--mismatch=<TEST> <GOLD> print lines with specified tag mismatches
--mismatches print all mismatches
--matrix print confusion matrix
--labeled evaluate syntax tiers with dependency labels
--unlabeled evaluate syntax tiers without dependency labels
--omit_single_words exclude single word utts from evaluation
dic is a bash script for looking up words in the CLAN lexicon. It will search for the term with word boundaries on either side. Thus searching for “how” will find “how”, “how_come”, or “know+how”, but not “howl”, “howitzer”, or “shower”.
The default output is in the same format as it would be in a transcript , i.e. pos|word(+comp|word...). Adding the -f switch shows output in the format of grep, i.e. filename:word {[scat pos]([comp comp+comp])}.
You can add the -p switch to search for parts of words. Thus dic -p how will find “how”, “how_come”, “know+how” AND “howl”, “howitzer”, and “shower”.
| dic how: | Searches lexicon for “how” with word boundaries. Output is same as in CHAT transcripts. |
|---|---|
| dic -f how: | Searches lexicon for “how” with word boundaries. Output is same as if grep had been used. |
| dic -p how: | Searches lexicon for any word containing the pattern “how”. Output is same as in CHAT transcripts. |
| dic -p -f how: | Alternatively, dic -f -p how. Searches lexicon for any word containing the pattern “how”. Output is same as if grep had been used. |