Hi Issac,
The "Read Lines" stage will transform the input document completely into a number of new documents (one for each line read).
You would need to tag the individual documents AFTER the "Read Lines" stage in the pipeline.
Yes, if you need the parent file information to use for generating the tag content, these are maintained as HCI_parentUri, HCI_parentId, and HCI_parentDisplay on the newly generated "line" documents.
It's a good feature request for introducing a mechanism to force metadata to be added to expanded documents from a parent document.
-Ben