[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [tor-talk] [OT] rewriting a text



On Fri, May 17, 2013 at 10:54 PM, NoName <antispam06@xxxxxxx> wrote:

>
> Is there a method to recompose some text? [...] One way I can think of is
> the use of some translation tool. Write it in some other language than the
> target and use the translation tool to get an impersonal version.
>

I have indeed heard of folks using e.g. google translate to let it do a
translation from [original lang] -> [some other lang] and back to ->
[original lang]. No sources though - would be interesting to try and let
the Anonymouth or any other tools identify authorship of such pieces of
text (input excerpts of original texts authored by multiple people ->
double-google-translate one of them -> see if it is identifiable still by
any of those tools.)

I've also heard of it being possible to deduce the relative amount of
entropy in text compared to some counterpart (another, corresponding text)
- the idea being, a translation of an original piece will have >= entropy,
which is deducible by summing up the general usage frequencies of each of
the words in a given piece of text (have to have word frequency
dictionary/data of course). This sounds a bit flimsy and in any case naive,
but if one were to implement such a measurement tool, two pieces of text
could then be compared by the amount of information (well, 1/entropy) they
carry.

It might make sense to write a simple tool which, given an original input
text, substitutes some general-but-replaceable words with more unique /
less unique counterparts from a thesaurus. However, syntax (and style,
still) would remain an issue. I've seen forum posts on some boards where
deliberate simple / pseudo-creole English is used, supposedly (such being
the speculation there) with the help of google/whatever translate - several
passes of. Would be interesting to hear of any (more) actual
studies/research about this, though.
_______________________________________________
tor-talk mailing list
tor-talk@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk