Ant colony heuristic for user-contributed comments summarization

详细信息    查看全文

• 作者：Mohamed Atef Mosa^a ; ^{mohammedatefmosa@gmail.com} ; ^{m.atef@narss.sci.eg} ; Alaa Hamouda^b ; ^{alaa.hamouda@azhar.edu.eg} ; Mahmoud Marei^b ; ^{marie@azhar.edu.eg}
• 关键词：Comments summarization ; Social media ; Swarm intelligence ; Ant colony optimization ; Text mining
• 刊名：Knowledge-Based Systems
• 出版年：2017
• 出版时间：15 February 2017
• 年：2017
• 卷：118
• 期：Complete
• 页码：105-114
• 全文大小：2063 K
• 卷排序：118

文摘

User-contributed comments (UCC) are one of the signs of the social media. Due to the high popularity of social media, it becomes already exceedingly difficult to find the most relevant, interactive information for the users. The motivation behind this work is the fact that users may interest to get an efficacious brief understanding of comments without reading the entire comments. This paper opens up an unconventional field of comment's summarization predicated on Ant colony optimization mixed with Jensen–Shannon divergence (ACO-JSD). ACO-JSD is a proposed novel technique concerning the extraction the most interactive comments from the huge amount of concise comment's perspectives. This problem is unfastened utilizing ACO to generate the optimal solution. Moreover, the JSD model is employed to ensure a summary could capture the essence of the original comments. First, an acyclic semi-graph has been constructed under two constraints: (1) the longest comments will be isolated from the graph, (2) The more similarity between two comments, the greater the chance that mutual connectivity is eliminated. Next, a feasible solution is constructed to select the high-quality summarization. Finally, the proposed algorithm has been evaluated over a collection of Facebook posts with their associated comments and an excellent performance in comparison with traditional document summarization algorithms was obtained. Accordingly, the computational results show the efficiency of the proposed algorithm, as well as its ability to find a good summary that is guaranteed to be near-optimal.