Basic Principles of Cross-Lingual Models

When we start working across languages, we need to determine ways of measuring similarity of entities (such as a word, a phrase, a sentence or a link between entities) within each language and across languages. Modern approaches usually rely on the Vector Space Model (VSM), which uses a numerical vector $$x$$ of a specified dimensionality $$D$$ to represent any entity. The entities are similar when their vectors are not too far apart.