Abstract
We present a study on the design of multilingual Answer Sentence Selection(AS2) models, which are a core component of modern Question Answering (QA)systems. The main idea is to transfer data, created from one resource richlanguage, e.g., English, to other languages, less rich in terms of resources.The main findings of this paper are: (i) the training data for AS2 translatedinto a target language can be used to effectively fine-tune a Transformer-basedmodel for that language; (ii) one multilingual Transformer model it is enoughto rank answers in multiple languages; and (iii) mixed-language question/answerpairs can be used to fine-tune models to select answers from any language,where the input question is just in one language. This highly reduces thecomplexity and technical requirement of a multilingual QA system. Ourexperiments validate the findings above, showing a modest drop, at most 3%,with respect to the state-of-the-art English model.