The good thing in doing a learner translation based research is that we've got the source text as a reference or benchmark.
Do not attempt to single out translation errors without any reference to grammatical (and still many more aspects of) errors. But it is always a good idea to start with some specific research questions, say, the translation of the relative clause into the target language.
I am told that Shanghai Foreign Studies University is compiling a corpus that contains Chinese students' translations in parallel to English/Chinese source texts.