No changes in the language identification computation from v1.4.3. But a major update of the code base and including a fixed schema validator. Before v1.4.4 the validation just silently skipped validation errors, leading to empty language identification files on s3.