Data Directory used when running test_phrase_grammar.py

[YianZhang]

Hi Yikang and other Contributors,

Thank you for making public the source code! I am trying to reproduce your results, but I am not sure what path to use as the command line argument of test_phrase_grammar --data. I downloaded PTB data and I am currently using treebank_3/parsed/mrg as the data argument. It does not work.

The listings under treebank_3/parsed/mrg:
atis brown readme.mrg swbd wsj

The listings under treebank_3/parsed/mrg/wsj:

00 06 12 18 24
01 07 13 19 MERGE.LOG
02 08 14 20
03 09 15 21
04 10 16 22
05 11 17 23

Thank you for your time!
Ian

[yikangshen]

Hi Ian,
You need to copy the wsj folder to ~/nltk_data/corpora/ptb/WSJ.

[YianZhang]

Hi Yikang,

Thanks for the response! I figured that out. However, what is args.data in test_phrase_grammar used for?

Thanks,
Ian

[yikangshen]

It points to the dictionary that the model actually uses.

[YianZhang]

It points to the dictionary that the model actually uses.

Thanks for the response! Do you mean "directory" or "dictionary"?

Best,
Ian

[yikangshen]

Dictionary

[yikangshen]

While testing parsing F1, the model still needs to load dictionary from training corpus

[YianZhang]

Thanks for your prompt response!

After carefully checking your code, I believe the dictionary is loaded from a fixed path:

Ordered-Neurons/test_phrase_grammar.py

Lines 279 to 282 in 46d63cd

	fn = 'corpus.{}.data'.format(hashlib.md5('data/penn'.encode()).hexdigest())
	print('Loading cached dataset...')
	corpus = torch.load(fn)
	dictionary = corpus.dictionary

And args.data is used as the directory of the test data:

Ordered-Neurons/test_phrase_grammar.py

Line 293 in 46d63cd

corpus = data_ptb.Corpus(args.data)

Am I correct?

Thanks for your help again! It would be appreciated if you can also check the other issue of mine: #25. As far as I know, this problem also confuses other researchers.

Best,
Ian

[shawntan]

The code assumes you have the cached dataset in the directory, and it would be cached if the training script was run prior to test_phrase_grammar.py.

But yes, you are correct.

[YianZhang]

Thanks a lot!