The manually annotated BIRD dataset

[SolidLao]

Thanks for the interesting work!

From the paper, there is a statement "we manually annotated the BIRD dataset (979 non-eq pairs in all 1534 pairs) and computed the recall for each approach". Is this annotated BIRD dataset available in the repository? From a quick scan, I did not find it.

Best,
Jiale

[chency87]

see Github Action:

https://github.com/sfu-db/ParSEval/actions/runs/28278708772/job/83790286272

One thing to note: I removed some hardcoded handling for the BIRD dataset, such as age computation and special datetime handling. As a result, the NEQ count may be lower than 979.

[SolidLao]

Thanks for answering!

Sorry for any confusion, but I want to clarify that, I did not mean the execution results of ParSEval, which is already available in the repository. I mean the manually annotated BIRD dataset where we have lables that 979 pairs are NEQ in all 1534 pairs; not the results that ParSEval identified 784 NEQ pairs.

Is this manually annnotated dataset available? Thanks!