How much data is enough for predicting transfer languages?
hzh1-cl opened this issue · comments
Hi, I am doing cross-lingual dependency parsing of a Tungusic language and would like to use LangRank to predict transfer languages.
I found that with different size of input data, the predicted results were different. I am wondering, what is the proper amount of data I should use as input? Could you please give me some advice? Thank you!