摘要
提出一种融合BM25与文本特征的新闻摘要算法。首先使用BM25算法计算TextRank算法中的句子相似度,其次选择词频和句子位置作为文本特征,最后将文本特征的评分与TextRank的评分相加作为文本中句子的评分,对所有的句子按照评分降序排列,选择评分最高的几个句子作为摘要。使用ROUGE工具在NLPCC2015数据集上进行测试,结果表明该方法有较好的效果。
This paper presents a news summary algorithm that combines BM25 and text features.Firstly,we use the BM25 algorithm to calculate the sentence similarity in the TextRank algorithm,then select the word frequency and sentence position as the text features,and take the text feature score and the TextRank score as the final score of the sentence in the text.Finally,we sort all the sentences in descending order according to the final score,and select the sentences with the highest scores as the news summary.The test results on the dataset of NLPCC2015 using ROUGE tools show that this method has a better performance.
作者
李楠
陶宏才
LI Nan;TAO Hong-cai(School of Information Science & Technology, Southwest Jiaotong University, Chengdu 611756,China)
出处
《成都信息工程大学学报》
2018年第2期113-118,共6页
Journal of Chengdu University of Information Technology
基金
国家自然科学基金资助项目(61505168)