Suggesting -- AI: AAAI+IJCAI+AISTATS; DM: KDD+CIKM+ICDM; IR: WWW+WSDM+SIGIR

[zhaoweijie12]

Dear Colleagues,

This would be my personal suggestion on revising the categories of AI, by taking into consideration of the reality of various communities. The goal is to boost the healthy growth of each sub-areas of AI and promote the excellence in research.

Artificial Intelligence: AAAI, IJCAI, AISTATS
Computer Vision: CVPR, ICCV, ECCV
Data Mining: KDD, CIKM, ICDM
Machine Learning: ICML, NeurIPS, ICLR
Natural Language Processing: ACL, EMNLP, NAACL
Information Retrieval: WWW, WSDM SIGIR

Let me elaborate on some the suggested revisions.

AAAI and IJCAI have been long criticized for publishing a large-number of not-top-high quality papers on machine learning, NLP, and CV. The argument for ranking AAAI and IJCAI as the top-tier conferences is that they also accept many high-quality traditional AI papers. It would hurt large number of researchers if AAAI + IJCAI are taken out of the list as top-tier conferences. Therefore, a remedy is to add AISTATS to this category. The average quality of AISTATS papers is higher than the average quality of machine learning papers in AAAI+IJCAI. Another choice might be UAI, but in recent years, it seems that AISTATS has attracted a lot more submissions than UAI. For example, in 2022, the number of submission for AISTATS is 1685 and the number for UAI is 712.

Data mining is still a large community. Regardless of how other communities view data mining, the field will continue to exist for long time. If only two conferences can be picked for data mining, we probably want to consider CIKM (in addition to KDD the obvious choice). CIKM attracts substantially more submissions (about 80% more than ICDM) and there are quite a few known works from CIKM. To many DM researchers, SDM is also as good as ICDM.

Information retrieval: WSDM is as good as SIGIR in essentially all aspects. There is no reason not to include WSDM if SIGIR is there.

[animesh-garg]

@emeryberger It may be useful to open Discussions
many of these are not active code base issues, but rather open-ended discussions which may benefit from community input and back-and-forth.