My bachelor thesis focues on the topic of multilingual language models and zero-shot learning. There is an increasing amount of evidence that in cases with little or no data in a target language, training on a different language can yield surprisingly good results. However, currently there are no established guidelines for choosing the training (source) language. In attempt to solve this issue we thoroughly analyze a state-of-the-art multilingual model and try to determine what impacts good transfer between languages. As opposed to the majority of multilingual NLP literature, we don’t only train on English, but on a group of almost 30 languages. I show that looking at particular syntactic features is 2-4 times more helpful in predicting the performance than an aggregated syntactic similarity. It appears that the importance of syntactic features strongly differs depending on the downstream task – no single feature is a good performance predictor for all NLP tasks. As a result, one should not expect that for a target language L1 there is a single language L2 that is the best choice for any NLP task (for instance, for Bulgarian, the best source language is French on POS tagging, Russian on NER and Thai on NLI). The most important linguistic features affecting the transfer quality are analyzed using statistical and machine learning methods.

Full paper: TBD

Code: https://github.com/blazejdolicki/multilingual-analysis