Hi all,
Every work in OpenAlex now gets its topics from a new classifier. The topics themselves aren't changing: same 4,516 topics, same names and IDs, same subfields, fields and domains. So anything you've built on topic IDs keeps working. What changes is which works get put in each topic.
The new assignments match what most folks expect a topic to mean: one that fits the work's title and abstract. The biggest gains are on works the old classifier had little to go on: papers with no abstract, papers not in English, and datasets and other records. It also fixes some bugs, like the 20 million mostly non-English works that all landed in "Military Technology and Strategies".
It's an incremental change. Institutions' field profiles stay about 97% the same, and across 3,621 universities mean FWCI barely moves (the median university changes 4.5%).
If you report on topics, check your numbers. We've posted before-and-after profiles for the 109 institutions that support OpenAlex:
https://github.com/ourresearch/openalex-topic-classification/tree/main/v2/changes/institutions
And change-set files for every topic, subfield and field:
https://github.com/ourresearch/openalex-topic-classification/tree/main/v2/changes
Topics now cover every kind of work, not just papers. If you want to count papers only, filter by type (articles, reviews, books and chapters). A few topics now hold millions of dataset records.
Need the old topics? They're frozen and you can download them here:
https://github.com/ourresearch/openalex-topic-classification/releases/tag/v2.0.0
The old classifier is still public, and /text/topics?version=1 keeps it running until 13 January:
https://help.openalex.org/api/tag-aboutness/#the-previous-topic-classifier-version1
Snapshot users: please reload fully from the 14 October snapshot.
Also new: topics and keywords are now hooked together. Each keyword shows the topics it's most used in, and each topic shows its characteristic keywords. You can poke around in the aboutness viewer:
https://openalex.org/aboutness/viewer
The full story, with examples, is on the blog:
https://blog.openalex.org/same-topics-better-assignments/
Thanks to everyone who's sent us examples when something looked off. Lots of them became our test cases!
Jason