Hi Anton,
the perspective you suggest can certainly provide a different angle. We do have a number of hints, e.g. the words that occur as labels are more likely to be nouns.
A couple of years ago, David suggested You are not allowed to view links. Register or Login to view.: a task that could be a subset of your proposal.
In any case, exact reduplication of words does not disappear even if one considers a nomenclator. And then You are not allowed to view links. Register or Login to view.n often appears as similar to You are not allowed to view links. Register or Login to view. (in particular, the q- prefix is frequent in both). Also, there are elements that suggest that the ending of a word has an influence on the prefix of the following word (see You are not allowed to view links. Register or Login to view.). For these reasons, I find it difficult to think of a purely-nomenclator solution. But thinking of ways to approach translation more directly seems useful to me: similarly to David's idea, a first step could be identifying groups of words that behave similarly (there are several algorithms that do this kind of stuff).
the perspective you suggest can certainly provide a different angle. We do have a number of hints, e.g. the words that occur as labels are more likely to be nouns.
A couple of years ago, David suggested You are not allowed to view links. Register or Login to view.: a task that could be a subset of your proposal.
In any case, exact reduplication of words does not disappear even if one considers a nomenclator. And then You are not allowed to view links. Register or Login to view.n often appears as similar to You are not allowed to view links. Register or Login to view. (in particular, the q- prefix is frequent in both). Also, there are elements that suggest that the ending of a word has an influence on the prefix of the following word (see You are not allowed to view links. Register or Login to view.). For these reasons, I find it difficult to think of a purely-nomenclator solution. But thinking of ways to approach translation more directly seems useful to me: similarly to David's idea, a first step could be identifying groups of words that behave similarly (there are several algorithms that do this kind of stuff).