您的位置:首页  > 论文页面

基于翻译模型的引文推荐

发表时间:2013-02-28  浏览量:1098  下载量:300
全部作者: 陆炀,陈翀,何靖,单栋栋,闫宏飞
作者单位: 北京大学网络与信息系统研究所;北京师范大学管理学院
摘 要: 探讨了如何为作者或编辑自动找到恰当的书籍文献以支撑其撰写的内容,即所谓引文推荐。主要方法为将引文推荐问题置于文本检索的框架之下,并将翻译模型融入传统的经典语言模型之中,以填充2种异构语言之间的鸿沟。通过比较不同的平行语料构建方法、翻译保留词汇量、语言模型与翻译模型权重参数等一系列的实验发现,与当下最优秀的方法相比,翻译模型可以找到更好的备选引文,且比已有的方法提升100%至400%不等。
关 键 词: 信息科学与系统科学基础学科;信息检索;引文推荐;翻译模型
Title: Recommending citations based on translation model
Author: LU Yang, CHEN Chong, HE Jing, SHAN Dongdong, YAN Hongfei
Organization: Institute of Network Computing and Information Systems, Peking University; School of Management, Beijing Normal University
Abstract: This paper discusses how to find appropriate literature automatically for authors and editors, namely the so-called citation recommendation, to support the materials who is writing about. Through treating the problem of information retrieval as a retrieval task and merging the translation model into the traditional classical language model, the gap between two heterogeous languages was bridged. Through a series of experiments, such as different sources for building the parallel texts, the preserved number of words in translation model, the weight parameters of language model and translation model, we found that the translation model can find out much better candidates of citations, when compared with the best method at present, the number of candidates of citations increased by 100% to 400%.
Key words: basic subject of information science and system science; information retrieval; citation recommendation; translation model
发表期数: 2013年2月第4期
引用格式: 陆炀,陈翀,何靖,等. 基于翻译模型的引文推荐[J]. 中国科技论文在线精品论文,2013,6(4):371-376.
 
0 评论数 0
暂无评论
友情链接