location: Current position: Home >> Scientific Research >> Paper Publications

ENHANCING TEXT REPRESENTATION FOR CLASSIFICATION TASKS WITH SEMANTIC GRAPH STRUCTURES

Hits:

Indexed by:期刊论文

Date of Publication:2011-05-01

Journal:6th International Symposium on Management Engineering (ISME 2009)

Included Journals:SCIE、EI、CPCI-S、Scopus

Volume:7

Issue:5B,SI

Page Number:2689-2698

ISSN No.:1349-4198

Key Words:Text representation; Graph structure; Maximum common subgraph; Classification

Abstract:To represent the textual knowledge more expressively, a kind of semantic-based graph structure is proposed, in which more semantic and ordering information among terms as well as the structural information of the text are incorporated. Such model can be constructed by extracting representative terms from texts and their mutually semantic relationships. Afterward, it is represented as a graph, whose nodes are the selected terms and whose edges are the corresponding relationships respectively. Moreover, the weight is assigned to each edge so that the strength of relationship between two terms can be measured. Furthermore, for this weighted directed graph structure, a novel graph similarity algorithm is developed by extracting the maximum common subgraph between two concerned graphs, which can therefore be used to measure the distance between two graph structures, i.e., two texts, and further be applied to classification tasks. Finally, some experiments have been conducted with the Chinese benchmark corpus for validation. The experimental results have proved the better performance of the proposed textual knowledge representation model in terms of its precision and recall.

Pre One:Temporal Topic Chain Mining Method based on the Scientific Literature

Next One:Uncertain Data Cluetser based on DBSCAN