Topic Aspect-Oriented Summarization Via Group Selection

Hanyin Fang,Weiming Lu,Fei Wu,Yin Zhang,Xindi Shang,Jian Shao,Yueting Zhuang

DOI: https://doi.org/10.1016/j.neucom.2014.08.031

IF: 6

2014-01-01

Neurocomputing

Abstract:The summarization is desirable to efficiently apprehend the gist of the huge amount of data and becomes a significant challenge in many applications such as news article summarization and social media mining. Considering the summaries from multi-documents of one topic can describe various aspects of one given topic, this paper attempts to exploit appropriate priors to generate topic aspect-oriented summarization (abbreviated as TAOS). The underlying intuition of the proposed TAOS is that different topics can prefer different aspects and the different aspects can be represented by different preference of features(e.g., technical topic may prefer proper noun than sports topic). In order to materialize the intuition of TAOS, we first extract several groups of features according to topic factors, and then a group norm penalty (i.e., ℓ1/ℓ2 norm) and latent variables are utilized to select overlapping groups of features. We compare our proposed approach with some state-of-the-art methods on DUC2003, DUC2004 datasets for text summarization and NUS-Wide dataset for image summarization. The results show our method can generate meaningful summarization in terms of ROUGE and Jensen–Shannon Divergence metrics.