I recall I have read about a parser which you just have to feed some sample lines, for it to know how to parse some text.
It just determines the difference between two l
Probably you mean TemplateMaker, I haven't tried it yet, but it builds on well-researched longest-common-substring algorithms and thus should work reasonably... If you are interested in different (more complex) approaches, you can easily find a lot of material on Google Scholar using the query "wrapper induction" or "template induction".
Conceivably you might mean Reverend?