论文部分内容阅读
[目的/意义]学术文献引文推荐是指对于给定的学术文献,自动化地为其推荐合适的引文和参考文献。借助于引文推荐,用户可以在一定程度上提高撰写学术文献的效率,降低对重要相关文献的漏引。[方法/过程]分析国内外引文推荐研究的最新进展,阐述引文推荐问题的演化过程,从局部引文推荐和全局引文推荐等方面对引文推荐进行梳理,重点归纳文档相似性、主题模型、翻译模型、协同过滤和混合推荐等5种引文推荐常用方法,并总结引文推荐常用数据集和测评方法。[结果/结论]已有引文推荐研究的主要问题在于未考虑用户偏好的动态变化性及研究领域的综合性,在用户研究和实际应用方面仍有所欠缺;未来引文推荐的研究可运用语义化表达方法和自然语言生成技术,从基于上下文的引文推荐和跨语言引文推荐等方面进行展开。
[Purpose / Significance] Academic Literature Citation refers to the recommendation of suitable citations and references for a given academic literature. With the help of citation recommendation, users can improve the efficiency of writing academic documents to a certain extent and reduce the leakage of important related documents. [Methods / Processes] This paper analyzes the latest research progress of citation recommendation at home and abroad, expounds the evolution of citation recommendation problem, sorts out citation recommendation from the aspects of local citation recommendation and global citation recommendation, and focuses on document similarity, thematic model, translation model , Collaborative filtering and mixed recommendation five citations recommend common methods, and citation recommended data sets and evaluation methods commonly used. [Results / Conclusions] The main problem of citation recommendation is that the dynamic variability of users’ preferences and the comprehensiveness of research fields are not considered, and there is still a lack of user research and practical application. Future citation recommended studies can use semanticization Expression methods and natural language generation techniques, starting from the context-based citation recommendation and cross-language citation recommendation.