Learning Kernels for Semantic Clustering: A Deep Approach

Abstract

In this thesis proposal we present a novel semantic embedding method, which aims at consistently performing semantic clustering at sentence level. Taking into account special aspects of Vector Space Models (VSMs), we propose to learn reproducing kernels in classification tasks. By this way, capturing spectral features from data is possible. These features make it theoretically plausible to model semantic similarity criteria in Hilbert spaces, i.e. the embedding spaces. We could improve the semantic assessment over embeddings, which are criterion-derived representations from traditional semantic vectors. The learned kernel could be easily transferred to clustering methods, where the Multi-Class Imbalance Problem is considered (e.g. semantic clustering of definitions of terms).

Unsupervised sentence representations as word information series: Revisiting TF–IDF

“…The main advantage of such an approach is that there exists the possibility of studying the statistical behavior of sentence meaning. As an additional and important benefit, sentence embeddings make it possible to leverage a number of NLP tasks, such as sentence clustering, text summarization (Zhang et al, 2012;Arroyo-Fernández, 2015;Arroyo-Fernández et al, 2016;Yu et al, 2017), sentence classification (Kalchbrenner et al, 2014;Chen et al, 2017;Er et al, 2016), paraphrase identification (Yin and Schütze, 2015), semantic similarity/ relatedness and sentiment classification (Arroyo-Fernández and Meza Ruiz, 2017;Chen et al, 2017;De Boom et al, 2016;Kalchbrenner et al, 2014;Onan et al, 2017;Yazdani and Popescu-Belis, 2013).…”