How Google generates titles for documents with language or script misalignment

Recently we introduced an algorithmic improvement that identifies documents where the title
element is written in a different language or script from its content, and chooses a title
that is similar to the language and script of the document.