Where do test dirs props, null, and ne come from?

[Adamits]

Hi!

I noticed in make-wsj-test.sh and make-brown-test.sh that we try to zcat a props, null, and ne file from test.wsj. However, in the extract_test_from_ptb.sh and extract_test_from_brown.sh scripts, none of these dirs/files are generated. Where are these supposed to come from?

Thanks!

[strubell]

Those dirs should be under the `train` directory in the conll05 data.

…

On Fri, Jul 9, 2021 at 6:49 PM Adam ***@***.***> wrote: Hi! I noticed in make-wsj-test.sh and make-brown-test.sh that we try to zcat a props, null, and ne file from test.wsj. However, in the extract_test_from_ptb.sh and extract_test_from_brown.sh scripts, none of these dirs/files are generated. Where are these supposed to come from? Thanks! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#9>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAY5TNYMBDKIEH4SM47TE6TTW54HZANCNFSM5ADRYS7A> .

[Adamits]

Thanks for the response!

I am probably missing something, but I thought the train directory only had data for sections 02-21 for wsj, whereas the test set is for sections 23. To be sure, I am referencing e.g. this line: https://github.com/strubell/preprocess-conll05/blob/master/bin/basic/make-wsj-test.sh#L13 - whereas https://github.com/strubell/preprocess-conll05/blob/master/bin/basic/extract_test_from_ptb.sh only generates words/syntax for section 23.

[XueBingo]

Hi!

I noticed in make-wsj-test.sh and make-brown-test.sh that we try to zcat a props, null, and ne file from test.wsj. However, in the extract_test_from_ptb.sh and extract_test_from_brown.sh scripts, none of these dirs/files are generated. Where are these supposed to come from?

Thanks!

Hello, I have the same problem with you.
Do you have any ideas now?
Thanks!

[strubell]

It sounds like you're describing the ptb training data, not the conll data - the directory I'm referring to is the $CONLL05 dir as defined in get_data.sh.

[Adamits]

Yeah I guess so. I am asking about the test data in particular. Which appears to be section 23 of PTB.

So running ./bin/basic/extract_test_from_ptb.sh only extracts words and synts from section 23.

However, bin/basic/make-wsj-test.sh expects props, null, and ne as well. I think for the train/dev data, these dirs come from the conll05 releaser, in get_data.sh, however, section 23 (the test data) does not seem to be included in here.

But for the test data, where do these dirs come from? In bin/basic/make-wsj-test.sh:

zcat < $CONLL05/$FILE/words/$FILE.words.gz > /tmp/$$.words
    zcat < $CONLL05/$FILE/props/$FILE.props.gz > /tmp/$$.props
    zcat < $CONLL05/$FILE/synt/$FILE.$s.synt.gz > /tmp/$$.synt

    # no senses, set to null
    zcat < $CONLL05/$FILE/null/$FILE.null.gz > /tmp/$$.senses
    zcat < $CONLL05/$FILE/ne/$FILE.ne.gz > /tmp/$$.ne

cannot find the props, sense, or ne file, and then writes an empty archive.

[strubell]

Oh, that's so strange! I guess the senses/ne lines (and corresponding entries in the paste) should be removed, but I'm surprised this non-working version is in the repo. Unfortunately I no longer have access to the old server where I originally developed/ran these scripts, so I can't go back and see if there were uncommitted changes, etc.