"Strings to document" node speed

"Strings to document" steal huge amount of time in processing data. And it seems like conversion which this node performs is more "cosmetical" operations than something needful. So my interest - is there way to "hack" common string data and convert it to "document" type without time loss?

The conversion from a document to a string can not be done completely without timeloss since a document is a more complex data structure than a simple string. A document is partitioned into paragraphs, sentences, terms, etc. Additionally when a document is created its text is tokenized and sentences are identified, which takes time as well.

How many strings do You want to convert into documents? I'm wondering since Your writing about "huge amount of time in processing data".