Concept Map Information Content Enhancement Using Joint Word Embedding and Latent Document Structure

Abstract

The concept map (CM) can be enhanced by extracting precise propositions, representing compactly, adding useful features that increase the information content (IC). To enhance the IC with domain knowledge of the document, an automatic enhanced CM generation using word embedding based concept and relation representation along with organization using latent semantic structure is proposed. To improve the concept significance, precise identification of similar items, clustering topically associated concepts, and hierarchical clustering of semantically related concepts are carried out. This augments the IC of the CM with additional information and generates CM with concise and informative content. The joint word embedding based on various contexts is utilized to determine distributional features critical for these enhancements. Summarization of the ECM to visualize the document summary is used to illustrate its resourcefulness. The work is evaluated using ACL anthology, Genia, and CRAFT dataset, and the information gain is approximately three times more in comparison with general CM.

Understanding Universal Adversarial Attack and Defense on Graph

“…The obtained graph representations can be applied to various downstream tasks, such as node classification , graph classification (Xie & Ying, 2021;Zhang et al, 2019) and community discovery (Chen et al, 2019;Zhang et al, 2020;Zhang et al, 2019). In addition, abstract contents, including images (Nhi et al, 2022) and documents (Stylianou et al, 2022;Ismail et al, 2022;Urkalan & Geetha, 2020), can be interpreted as nodes in the graph. The graph-based methods help discovering the relationship among contents.…”