Feature Request: support for CoNLL format about croissant

[benjelloun]

Hi,

CoNLL uses at TSV format for its files. This can be described in Croissant similarly to the CSV examples, e.g., :

https://github.com/mlcommons/croissant/blob/main/datasets/1.0/movielens/metadata.json

For the data files, you need to specify: encoding_format="text/tab-separated-values" on FileObject.

The Croissant editor should figure things out automatically if you upload the data files one by one.

Best,
Omar

[gzhang64]

Thanks Omar for the suggestion!

To my best knowledge, CoNLL is not a strictly TSV file (yes, it's indeed tab separated):

1. CoNLL does not necessarily have column names, this is the major problem I have after uploading the file into Croissant Editor. I will have to manually add column names, but it's unclear how the added names will be associated with the columns.
2. CoNLL may have empty lines to as separators between sentences, while in regular TSV files, empty strings are interpreted as empty fields. I am not sure how empty lines in CoNLL files are parsed.