Jump to content

Web2Cit/Research/Automatic translation tests

From Meta, a Wikimedia project coordination wiki

Translation tests represent Web2Cit's expected output for specific target webpages, so the citation metadata in these tests must be correct and accurate. We used the citation metadata extracted from featured articles citations to build 80 translation tests files for the most cited domains having low scores. We selected domains cited more than 100 times in our corpus and having correct information 3 or less (more details about domain selection and file creation in create-translation-tests.ipynb). The files generated for the following domains were excluded from upload since tests previously created by users already existed: www.amazon.com, abcnews.go.com, edition.cnn.com, www1.folha.uol.com.br.

The following table presents the automatically created translation tests we added to Web2Cit:

Translation tests for highly-cited low-performing domains

[edit]