Linear transformations for cross-lingual semantic textual similarity

摘要: Abstract Cross-lingual semantic textual similarity systems estimate the degree of meaning between two sentences, each in a different language. State-of-the-art algorithms usually employ machine translation and combine vast amount features, making approach strongly supervised, resource rich, difficult to use for poorly-resourced languages. In this paper, we study linear transformations, which project monolingual spaces into shared space using bilingual dictionaries. We propose novel transformation, builds on best ideas from prior works. experiment with unsupervised techniques sentence based only show they can be significantly improved by word weighting. Our transformation outperforms other methods together weighting leads very promising results several datasets