Open Source Summit + ELC North America 2019

To comply with open source licenses, users must include proper license text in their software. Unfortunately, many times they include poorly-worded texts instead. To help solve this problem, Microsoft is leading an effort to clearly define license statements at clearlydefined.io.

In order to facilitate this effort, the Canvass Labs team analyzed the entire corpus of license text in Linux and Android and grouped each license according to a license type defined by The Software Package Data Exchange (SPDX). Then, they cataloged how each instance appearing in the Linux and Android repository varied from the standard license template.

They found that the license variations formed groups sharing similar structure for each underlying license template. They also found that sentence-level granularity leads to intuitive grouping.