Most popular scholarly works in the English Wikipedia and their transition to open access

Following the release of "The future of OA" by Piwowar, Priem, Orr (2019), interest has grown on how to accelerate the share of scholarly works consultations which meet an open access record.

Based on download patterns for over 23 million DOIs in 2017, released by Elbakyan (2018), we found that the 1 million most downloaded DOIs accounted for over 30 % of the total downloads. Of these 1 million DOIs, over 50 thousands (5 %) were previously identified as cited on the English Wikipedia and not open access (Leva 2018). Of these, 2440 DOIs are now open access according to the Unpaywall API as of 2019-10-25: a list of the corresponding OA URL and host type is enclosed, showing that 34 % became OA at the publisher while 66 % were made OA by a repository. The newly OA works were hosted at over 400 domains of which over 300 repositories, but the top 10 repositories accounted for a large portion of the works, with the top 3 repositories accounting for over 40 % of the newly found green open access DOIs.

Part of the newly OA works were just false negatives in Unpaywall in 2018, but a small manual sample shows that most are truly new deposits. Works from 2017 can be expected to be over-represented in the sample given that they were probably the most popular downloads of 2017 and could have been under embargo in 2018 when the previous measure of open access status was made.

自Piwowar、Priem与Orr于2019年发布《开放获取的未来》（*The future of OA*）一文以来，学界对于如何提升符合开放获取（Open Access）标准的学术作品访问量占比的关注度与日俱增。

依托Elbakyan于2018年发布的2017年2300余万个数字对象标识符（Digital Object Identifier, DOI）下载模式数据，本研究发现下载量排名前100万的DOI占总下载量的30%以上。在这100万个DOI中，有超过5万篇（占比5%）此前被标注为在英文维基百科中被引用且未开放获取（Leva, 2018）。截至2019年10月25日，根据Unpaywall API的数据，其中2440个DOI现已实现开放获取：随文附上了对应开放获取URL与托管类型的清单，结果显示34%的作品通过出版方渠道实现开放获取，剩余66%则通过学术仓储完成开放获取。

这批新实现开放获取的作品托管于超过400个域名，其中包含300余个学术仓储，但前10大仓储承载了其中绝大多数作品，而排名前三的仓储则涵盖了新发现的绿色开放获取（Green Open Access）DOI总量的40%以上。

部分新开放获取的作品仅为2018年Unpaywall数据库中的假阴性结果，但小规模人工抽样显示，其中绝大多数实为新增的仓储存档。由于2017年的作品大概率是2017年下载量最高的学术资源，且在2018年（彼时此前的开放获取状态评估已完成）可能仍处于出版禁运期，因此2017年的作品在样本中的占比可能偏高。