Handing of hetronyms

[JCGoran]

Disclaimer: I'm not a linguist, so maybe this is a dumb question. How are heteronyms handled in epitran? I haven't found any references to them in the repo or the paper. For instance, this code:

epi = epitran.Epitran('eng-Latn')
print(epi.transliterate('live'))

outputs lajv, so I'm curious how does it decide to use this pronunciation instead of lɪv.