We currently have two internationalization guides:
The first is the way we always did things up to the point where GSoC2012 finished. The second is the way to use only the GSoC results. However, the boundaries between the two are not clean cut. You could, for example, use spotters from the old code with disambiguators from the new code. It is also not clear yet to us which code is definitely better. There is a long list of questions without answers (need more experimentation), although the new code has already been shown to be more memory and time efficient if you are looking mostly for named entities.
We will rework these pipelines over the next months so as to produce a better organized guide and more streamlined process.
DBpedia Spotlight - Shedding Light on the Web of Documents