About ACM Dataset

[MavenZheng1003]

Dear Professor/Expert,

I hope this message finds you well. I am writing to seek your guidance regarding the two versions of the ACM dataset (the 3025-node and 4019-node versions). According to the literature, the additional nodes in the 4019-node version are primarily from KDD-related papers. However, when I manually supplemented the 3025-node version with KDD papers to expand it to 4019 nodes, the experimental results exhibited significant discrepancies compared to the official 4019-node version.

To better understand this inconsistency, I would like to kindly ask the following questions:

How was the official 4019-node version of the ACM dataset constructed? Specifically, what criteria were used to select the additional nodes?

Is the raw textual information (e.g., paper titles, abstracts, or metadata) corresponding to each node in this version publicly available?

This information would greatly assist me in comprehending the dataset construction methodology and reproducing the experimental results. Thank you very much for your time and support.

[Theheavens]

The toy ACM dataset is too old and perhaps no one remembers how it was originally constructed.
Maybe you could refer to https://snap-stanford.github.io/ogb-web/ or https://dl.acm.org/doi/10.1145/3583780.3615117 for criteria of dataset construction.