10k is quite a bit too much.
All words (word types) that occur only once in the MS are by definition part of only one language just by chance.
Similarly, all words that occur only twice may end up in the same language just by chance, at a 50% probability.
(This is simplified, based on a model that half the text is language A and the other half is language B. Therefore, the numbers are qualitative, but not far off).
For all words that occur only three times in the MS, the probability that they are all in the same language just by chance is 25%.
One may decide where to put the limit, but one clearly should look only at words that occur at least 3 or 4, perhaps even more times. That greatly reduces the number of word types that should be used for such an analysis.
All words (word types) that occur only once in the MS are by definition part of only one language just by chance.
Similarly, all words that occur only twice may end up in the same language just by chance, at a 50% probability.
(This is simplified, based on a model that half the text is language A and the other half is language B. Therefore, the numbers are qualitative, but not far off).
For all words that occur only three times in the MS, the probability that they are all in the same language just by chance is 25%.
One may decide where to put the limit, but one clearly should look only at words that occur at least 3 or 4, perhaps even more times. That greatly reduces the number of word types that should be used for such an analysis.