Skip to content

add fra tasks#11

Open
AleksandrAstakhov wants to merge 3 commits into
dialogmtebfrom
add_fra_tasks
Open

add fra tasks#11
AleksandrAstakhov wants to merge 3 commits into
dialogmtebfrom
add_fra_tasks

Conversation

@AleksandrAstakhov

Copy link
Copy Markdown
Collaborator

Add French MTEB tasks for the DeepPavlov dialogue/intent datasets that had no fra subset yet: ViraIntent, Clinc150, DailyDialog (act/emotion), AirDialogue, Mantis, AbgCosQA, ClarQA, Canard, WizardOfWikipedia, Coral, FaithDial and TopiOCQA.

If you add a model or a dataset, please add the corresponding checklist:

Add French MTEB tasks for the DeepPavlov dialogue/intent datasets that had
no fra subset yet: ViraIntent, Clinc150, DailyDialog (act/emotion),
AirDialogue, Mantis, AbgCosQA, ClarQA, Canard, WizardOfWikipedia, Coral,
FaithDial and TopiOCQA.

@Samoed Samoed left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

tmp

Comment on lines +31 to +38
bibtex_citation=r"""
@inproceedings{guo2021abg,
author = {Guo, Meiqi and Zhang, Mingda and Reddy, Siva and Alikhani, Malihe},
booktitle = {3rd Conference on Automated Knowledge Base Construction},
title = {Abg-coqa: Clarifying ambiguity in conversational question answering},
year = {2021},
}
""",

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Лучше пока не добавлять цитаты, тк это мы сделали перевод

Fill in missing required TaskMetadata fields (reference, date, license)
that caused test_all_metadata_is_filled_and_valid to fail for
FrAbgCosQA, FrClarQA, FrCanard, FrMantisClassification and
FrWizardOfWikipedia, using each task's own DeepPavlov dataset page as
reference instead of the original English dataset's.

Clear bibtex_citation on all fra tasks since these are DeepPavlov's
French translations, not the cited work itself. Also fixes FrCoral's
reference, which pointed at an unrelated audio dataset, and
FrWizardOfWikipedia's bibtex, which was a copy-paste of Canard's.
Add FrWebLINXCandidatesReranking for DeepPavlov/weblinx_fr, the last
available (non-blocked) French dataset from the extended dataset list.
Uses AbsTaskReranking like RuWebLINXCandidatesReranking; no
dataset_transform needed since the "query" column already holds the
French text directly (query_dict keeps the original as a side field).
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants