Unsupervised Neural Aspect Extraction with Sememes

Abstract: Aspect extraction relies on identifying aspects by discovering coherence among words, which is chal- lenging when word meanings are diversified and processing on short texts. To enhance the perfor- mance on aspect extraction, leveraging lexical se- mantic resources is a possible solution to such chal- lenge. In this paper, we present an unsupervised neural framework that leverages sememes to en- hance lexical semantics. The overall framework is analogous to an autoenoder which reconstructs sen- tence representations and learns aspects by latent variables. Two models that form sentence repre- sentations are proposed by exploiting sememes via (1) a hierarchical attention; (2) a context-enhanced attention. Experiments on two real-world datasets demonstrate the validity and the effectiveness of our models, which significantly outperforms exist- ing baselines.

Download: [PDF]