Note: I interpret occurrence rate as being the primary form of weight (aka value) determination for a word.
Note: These are probably genre-biased, in the sense that most of my suggestions like footnote referencing are more important to technical writing than to easy-reading (read thin) fiction.
The following things I think would be extremely cool and useful from (lib)ots.  I will spend a little bit of time on them at least, but if you would like to help implement something in here, you are welcome and encouraged to, but please email me (my name without spaces at hotpop dot com) so there isn't any wasteful duplication of work.

1) When encountering a capitalized word, or word containing capitals, search through the end of the sentence for parenthetical notation of an acronym, and if it exists, add weight to the sentence equal to .5 times the occurrence rate of the acronym.  When encountering the full name again, replace it with the acronym.  Add weight to the acronym itself equal to .1 times the acronym's occurrence rate, because a sentence containing the acronym obviously relates to the acronym, and the more the acronym occurs,  the more important it is to the subject of the paper.
2) When encountering a foot/end/margin note reference in the form of [#] where # is an integer, numeral, or other common symbol for this use, depending on the argument setting to OTS, give a little more weight to the note, and find a good place to put it if it qualifies to be present in the summary, or simply place it where it was (at the end for foot- and endnotes, margin notes are tricky and would have to be placed within the summary if qualifying, unless OTS rewrites *note references to correspond more to the summary/abstract, which is probably a good idea, especially for longer docs).
3) Provide various templates for Automated Abstract Creation (to be implemented).  Focus on common, convertable file types (sgml), in addition to plain text.
4) Interpret wildcards.  Such as recognizing that *note probably refers to [footnote, endnote, marginnote, /foot\-.*/, ...]

Other stuff:
German dictionary - shouldn't be hard at all
Various other languages...even if we don't know them, can't we just run ots on an extensive variety of works in that language, and create a 'probably works'  dic and check it via translation?  Shouldn't be hard.
The next step, of course, is finding or making an extremely powerful translator that makes few if any grammatical errors and then you can have an automated web archive for submitting a paper in language X and having it generate and optionally publish the abstract for the doc in language X, as well as the doc in languages A, B, C, D, and E, and abstracts for each of those.  Just something to think about.  Of course, the more people follow the guidelines in the ots english lesson (which should be renamed writing lesson), the more effective such a system could be.
The only tips in the english lesson that dont apply to writing in general (writing not geared towards summarization) are the ones that pertain to optional ots features, so you could opt to turn those off if the fact that your paper wasnt written for ots was causing a suboptimal summary production.  Even better, if such variances are consistent, if it is generic enough, patch ots to accept it as a new scheme, otherwise have the ability to make an otsrc with multiple preference schemes (--use-style=mystyle3, --conf=$HOME/mybooks/.otsrc) and customizability of the methods in ots.
