How do machines translate linguistically distant languages?

 In current times, machine translation is key for intercultural communication, particularly in multimedia platforms. A common experience is that machine translation works well for typologically-related, lexically-syntactically similar language pairs (e.g. Norwegian to Swedish) but not so much in a typologically distant pair (e.g. English to Japanese). So how are scientists working towards improving the latter?

This critique, by Yan-Yi Lee for Varsity, can be found here. (Note that this is one of the editor's old articles in another journalistic outlet)