Towards a Computationally Relevant Typology for Polyglot/Multilingual NLP

The purpose of this writing is to facilitate discussion on possible future directions for typology for polyglot/multilingual NLP. Although there is a shared objective between linguistics and polyglot NLP of accounting for all languages (or as many as possible) using one framework, there are many differences that we need to take into consideration. We identify some of these differences and their underlying reasons, and provide some suggestions as to how we can build a computationally relevant typology that could be useful for both the science and engineering of language.