Each relation should reference the original text with citation info

[Quantisan]

we can do this if we keep track of the original text and citation throughout the entire data extraction pipeline. Each distilled text should carry the reference to original text.