In the previous snapshot, “Using PARAT to Rank Companies by Top AI Conference Publications,” we explored the companies in PARAT that rank highest in AI research productivity. We specifically examined the top five companies, both in terms of the companies’ overall AI research publication count and in percentage of AI research published at top conferences. Here, we explore the top AI conference venues, and the distribution of publications across these venues, for the 705 companies in PARAT that have AI publications.
For top AI conferences in PARAT, we use 13 conferences that have broad calls for papers in general AI and machine learning research fields, as well as more focused conferences in areas like natural language processing and computer vision. Because the focus of the conferences differ, it may be telling to see where companies are publishing their AI-related research.
Below is the list of top AI conferences used to generate the total publication counts for the Top Conference AI Publications metric:1
- AAAI Conference on Artificial Intelligence (AAAI)
- International Joint Conference on Artificial Intelligence (IJCAI)
- IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
- European Conference on Computer Vision (ECCV)
- IEEE International Conference on Computer Vision (ICCV)
- International Conference on Machine Learning (ICML)
- International Conference on Knowledge Discovery Data Mining (SIGKDD)
- Conference on Neural Information Processing Systems (NeurIPS)
- Annual Meeting of the Association for Computational Linguistics (ACL)
- North American Chapter of the Association for Computational Linguistics (NAACL)
- Conference on Empirical Methods in Natural Language Processing (EMNLP)
- International Conference on Research and Development in Information Retrieval (SIGIR)
- The Web Conference (WWW)
For each of the top 13 AI conferences used in PARAT, we count the number of company publications between 2010 and 2020 (see Figure 1). We see that IJCAI (5,457 publications), AAAI (5,329 publications), and ICML (3,793 publications) are the top three conferences by company publication count in this 10-year span. IJCAI, AAAI, and ICML have broad calls for papers in a wide range of AI and machine learning areas. Conferences like ICCV and CVPR that are focused on one area of AI/ML, specifically computer vision, have fewer company publications over the past decade.
Figure 1. Number of AI publications with PARAT company-affiliated authors, by conference 2010-2020
Source: CSET’s Private-sector AI-Related Activity Tracker and Merged Corpus
These conferences differ not only in their focus, but also in their selection criteria and acceptance rates. To understand how competitive these conferences are, we provide details in Table 1 on the publication submission and acceptance rates for the top three (IJCAI, AAAI, and ICML) and bottom three (SIGKDD, CVPR, ICCV) most published in 2020 AI conferences.2 IJCAI had the lowest acceptance rate, at 13 percent, while AAAI and ICML had acceptance rates at 21 and 22 percent, respectively. CVPR and ICCV have the highest acceptance rates at 25 percent each.
Table 1. Select top AI conference 2020 publication submission statistics.
|Conference Name||Number of Submissions||Number of Acceptances||Acceptance Rate|
From 2015 to 2020, there was a shift among the top three conferences in where PARAT companies were publishing. While IJCAI lead in publications overall (as shown in Figure 1), it did not maintain its lead in 2019 and 2020. In 2019, AAAI surpassed IJCAI with 866 publications. In 2020, ICML surpassed IJCAI with 1,155 publications.
Figure 2. Number of company publications from 2015 to 2020 in the top five most published in AI conferences
Source: CSET’s Private-sector AI-Related Activity Tracker
After exploring AI publications in PARAT, our next Snapshots will look into more details about companies’ AI patenting activity.
- This list is derived from http://csrankings.org/
- ICCV’s data is for the 2019 conference.