... the last 10 sampling iterations. We also set a threshold to elesk similarity values, which yieldsbetter performance. Same as (Sinha and Mihalcea,2007), values of elesk larger than 240 are set ... senses acrossdictionaries, hence Wik is only used as augmenteddata for WMF to better learn the semantics of words.All data is tokenized, POS tagged (Toutanova et al.,2003) and lemmatized, ... challenge lies in that sense defini-tions are typically too short/sparse for latent vari-able models to learn accurate semantics, since thesemodels are designed for long documents. For exam-ple, topic...