Wikipedia-Graph Based Key Concept Extraction towards News Analysis
Baoyao Zhou, Ping Luo, Yuhong Xiong, Wei Liu · 2009
The well-known Wikipedia can serve as a comprehensive knowledge repository to facilitate textual content analysis, due to its abundance, high quality and well-structuring. In this paper, we propose WikiRank - a Wikipedia-graph based ranking model, which can be used to extract key Wikipedia concepts from a document. These key concepts can be regarded as the most salient terms to represent the theme of the document. Different from other existing graph-based ranking algorithms, the concept graph used for ranking in this model is constructed by leveraging not only the co-occurrence relations within the local context of a document but also the preprocessed hyperlink-structure of Wikipedia. We have applied the proposed WikiRank model with the Support Propagation ranking algorithm to analyze the news articles, especially for enterprise news. These promising applications include Wikipedia Concept Linking and Enterprise Concept Cloud Generation.