Conference size statistics over time


These are the number of paper submissions and acceptances for various CS-related conferences over time, including from AI, ML, HCI, NLP, Vision, Databases, and Data Mining, Web, and IR. Y-axis is a log scale, so acceptance rate corresponds to the vertical space between dots.

Fun notes:

The biggest caveat is that different data sources may be counting different things. The intention was to only count acceptances for "full" (as opposed to short papers) or "main session" (as opposed to demos or workshops) or "research track" (as opposed to industry track) papers. Different places have different definitions and standards for these things; for example, sometimes full and short have different submission tracks, or sometimes they're combined and there are conditional acceptances as different types, and sometimes this correlates to poster-or-not, etc. And these standards change over time (for some venues, papers in the 1980s are often quite a bit shorter than ones today). Ideally, the counts here would be for what each research community internally considers to be reasonably high-status publications representing a substantial amount of work. But I am not familiar with the internal conventions or status designation norms within most of the communities here, and am hoping the data sources are a reasonable proxy for that.

Some of the data sources for different conferences include:

The spreadsheet has more precise information on the sourcing, including URLs and attempts to document every data point, and listing multiple contradictory data sources when they exist. Updates and fixes welcome.

Link: Data and plots

Posted March 2018 by Brendan