Online Cross-Lingual PLSI for Evolutionary Theme Patterns Analysis.

In this paper, we focus on the problem of evolutionary theme patterns (ETP) analysis in cross-lingual scenarios. Previously, cross-lingual topic models in batch mode have been explored. By directly applying such techniques in ETP analysis, however, two limitations would arise. (1) It is time-consuming to re-train all the latent themes for each time interval in the time sequence. (2) The latent themes between two adjacent time intervals might lose continuity. This motivates us to utilize online algorithms to solve these limitations. The research of online topic models is not novel, but previous work cannot be directly employed, because they mainly target at monolingual texts. Consequently, we propose an online cross-lingual topic model. By experimental verification in a real world dataset, we demonstrate that our algorithm performs well in the ETP analysis task. It can efficiently reduce the updating time complexity; and it is effective in solving the continuity limitation. © Springer-Verlag 2013.