Cognitive memory-inspired sentence ordering model

Publication Type:
Journal Article
Knowledge-Based Systems, 2016, 104 pp. 1 - 13
Issue Date:
Filename Description Size
1-s2.0-S0950705116300326-main.pdfPublished Version1.4 MB
Adobe PDF
Full metadata record
As the novel web social media emerges on the web, large scale unordered sentences are springing up in the forms: news headlines, microblogs, comments and so on. Although these massive sentences contain rich information, their loose semantic association and highly unordered semantic organization make web users extremely difficult to capture the rich information due to the lack of semantic coherence. Sentence ordering is a significant research area focusing on obtaining coherent sentence orders which could assist web user to easily understand these unordered sentences. Although many state-of-the-art coherence-based sentence ordering approaches have been proposed, some challenging issues remain unsolved: 1) what is knowledge foundation to support semantic representation and inference in sentence ordering; 2) what are the cognitive mechanisms used to guide sentence ordering; 3) how to collaborate different cognitive mechanisms for generating a coherent sentence order. To solve these issues, we propose a cognitive memory-inspired sentence ordering model. We propose three cognitive logical structures and construct their corresponding markov random fields to support semantic representation and inference. Besides, we propose three memory-inspired modules and their corresponding learning methods. These modules collaborate with each other to generate a coherent sentence order. Comparative experiments are conducted to evaluate the effectiveness of the proposed model. The results show our model can obtain coherent sentence orders with higher accuracy compared with the baseline methods.
Please use this identifier to cite or link to this item: