Measuring Semantic Relatedness using Multilingual Representations
Samer Hassan, Carmen Banea, Rada F. Mihalcea · 2012
This paper explores the hypothesis that semantic relatedness may be more reliably inferred by using a multilingual space, as compared to the typical monolingual representation. Through evaluations using several stateof-the-art semantic relatedness systems, applied on standard datasets, we show that a multilingual approach is better suited for this task, and leads to improvements of up to 47% with respect to the monolingual baseline. 1