The mistake: wrong name for the train set and dev set for UP_Portuguese-Bosque

[scofield7419]

You may wrongly named the train set file and dev set file for UP_Portuguese-Bosque by mistake.
The two file names need to be exchanged.

[arademaker]

This issue is solved but many problems were found during the merge of the current UD data from the master branch of http://github.com/universaldependencies/UD_Portuguese-Bosque and the SRL annotations in the https://github.com/System-T/UniversalPropositions/tree/master/UP_Portuguese-Bosque.

See https://github.com/System-T/UniversalPropositions/blob/issue-4/UP_Portuguese-Bosque/README.md#merging-up-and-ud-datasets

We will need to manually inspect some sentences and decide about the final format of the files in this repository. To make Portuguese data similar to the remain files, I will temporarily convert the new CoNLL-U valid files to CoNLL files (one extra column for each predicate).

[arademaker]

@huaiyu-zhu comments are welcome.

[arademaker]

I found at least one sentence in the current set of PT sentences in the MASTER branch with a wrong sent_id. Fixed in 109816f

[yunyaoli]

@arademaker Can we close this issue? Thanks.

[alanakbik]

Hello @arademaker are you planning to merge branch issue-4? We just stumbled across the problems in the Portuguese UP and it seems to fix them?

[arademaker]

Yes, I need to finish this issue. I will need 1-2 weeks for that.

[yunyaoli]

@arademaker Any updates?

[arademaker]

Sorry, I will solve it this week. But we still need to reevaluate the plans for the PR #7.

[andreabac3]

UP

[arademaker]

Sorry, can you elaborate?

[andreabac3]

@arademaker Sorry, I'm referring to the train and dev file naming problem. I only realized this after producing the dataset statistics. With my comment ("UP"), I wanted to bring to attention the fact that the names are still reversed.

I take this opportunity to thank all of you for your work.

Kind regards,
Andrea