OpenAlex API features, improvements, and changes
Casey Meyer
We've been BUSY improving OpenAlex and I have several improvements to announce. We really appreciate all the feedback, so keep it coming!
Features
OpenAlex now has fulltext search across 57.3 million works, courtesy of n-grams (words, numbers, phrases in an article) provided by The General Index. We have a blog post coming soon, but here are some details.
Fulltext is already available via the main search parameter:
https://api.openalex.org/works?search=dna
With this method it receives less weight than title and abstract. But it's going to find words or phrases deep within articles, that title and abstract sometimes do not cover. Want to search only the works that contain fulltext? Go for it! You can also filter works to see only those that contain fulltext.
What about the n-grams themselves? Those are available too! Each work has an ngrams_url field that will lead you to n-grams if they exist for that work:
https://api.openalex.org/works/W2063450941/ngrams
You can look up n-grams by DOI as well:
https://api.openalex.org/works/10.1016/s0022-2836(75)80083-0/ngrams
This is a huge improvement for search, often boosting search results by 30x. Coverage across works is discussed here if you want to learn more.
Improvements
We have added these new convenience filters to works:
- The filter raw_affiliation_string.search now searches within strings, so queries such as “france” return records with "parisFrance".
- You can filter by ID fields in works, such as "openalex", which is useful for querying multiple IDs with the OR operator. We are adding this to the other entities soon.
- We fixed a bug in authors autocomplete that affected queries with two periods in a name.
- We fixed a bug to ensure sort order remains the same when searching and paging with cursors.
- The default sort order for works is now cited_by_count descending, rather than publication_date descending
- The author hint for the author autocomplete endpoint is now sorted by most cited rather than most recent. We're still working on this and will likely tweak it to prefer newer, highly-cited works.
