Right now, the ir-datasets.bib file is a bit messy, with inconsistencies in the ids/fields/formatting/etc. across records. It's probably best to go with an established source, such as DBLP, the ACL Anthology, the ACM DL, and/or the IR Anthology.
I'm inclined to go with DBLP when possible. I believe it covers most (all?) the papers we need, and has IDs that look nice. The IR Anthology also takes bibtex from here.
In cases where DBLP doesn't cover the paper, we can just mimic the look/fields/etc.
I don't think there's a need to pull the bibtex automatically -- I'm happy just to copy it over.
Right now, the
ir-datasets.bibfile is a bit messy, with inconsistencies in the ids/fields/formatting/etc. across records. It's probably best to go with an established source, such as DBLP, the ACL Anthology, the ACM DL, and/or the IR Anthology.I'm inclined to go with DBLP when possible. I believe it covers most (all?) the papers we need, and has IDs that look nice. The IR Anthology also takes bibtex from here.
In cases where DBLP doesn't cover the paper, we can just mimic the look/fields/etc.
I don't think there's a need to pull the bibtex automatically -- I'm happy just to copy it over.