About the data
This explorer is built on a sample of computer science research, not on all of it. Read every chart as “in this sample”.
- Source
- OpenAlex, field 17 (Computer Science), articles with an abstract of at least 200 characters
- Published
- January 2015 to October 2026, drawn evenly from every month
- Papers
- About 100,000
- Largest share
- China and the United States
What the sample does and does not cover
- Only papers OpenAlex files under Computer Science. Computer science work filed under other fields (medicine, physics, linguistics) is mostly missing, and some papers OpenAlex labels as computer science carry topics from other fields.
- Each month contributes a similar number of papers, so a raw count is flat by design. The Progress page shows the estimated real output per year.
- Country shares follow author affiliations in OpenAlex. Recent papers have fewer affiliations recorded, so the newest months under-count some countries.
- Hong Kong, Macao, Taiwan, Réunion and similar territories are shown as regions, as OpenAlex lists them separately.
What a “gap” means
A research gap here is a topic that grew fast but still has few papers in this sample, or a pair of topics that rarely appear together. It is not proof that nobody in the world works on it. Papers outside this sample can easily fill it.
Reading the galaxy
Each dot is a paper and nearby papers have similar meaning. The galaxy shows a sample of papers chosen to represent every cluster; clusters are named after their most distinctive topics.
Growth figures
Growth trends are hidden for now: only about 72% of 2026 papers have author countries recorded, against 97% of 2025 papers, so a year-to-date comparison would show false declines. When they return, growth compares two periods of equal length. When either period has fewer than 30 papers, it will show “Not enough data” instead of a percentage, because small counts swing wildly.
How search works today
Search and the galaxy layout use the BAAI bge-small sentence-embedding model (384 dimensions), so papers are matched by meaning, not only by shared words. Results below a similarity floor are dropped, so an off-topic query returns no match rather than a weak one.