可视辅助科学文献阅读的研究及应用

发布时间：2018-03-30 18:46

本文选题：文档可视化　切入点：文本摘要技术　出处：《天津大学》2016年硕士论文

【摘要】：近年,科技论文发表数与日俱增,研究学者需要阅读越来越多的文献。怎样快速有效地阅读一篇科技论文逐渐成为一个重要的研究问题。而一篇科技论文通常是一项学术研究的结晶,其中涵盖了许多了论点和发现。这使读者很难在很短的时间内获取到一篇文章的核心论点。另一方面,在阅读科技论文时,理解与其相关的重要参考文献对于更好的理解文章内容有很大的帮助。然而,一篇文章的引用网络是一个复杂的结构。在引用网络中寻找相关文献很容易会让人迷失在文献的复杂网络中。由此,怎样从众多的参考文献中快速找到最重要最相关的几篇,怎样能在阅读过程中不迷失在文档的多维空间中也成为值得研究的问题。本文利用文本可视分析的技术进行科学文献阅读的研究。为了解决上述问题,本文通过研究科学文献在引用网络中的关系和特性,在文档分析相关研究和技术的基础上,提出利用一种基于阅读目的的文本摘要技术来进行文章中关键句子的抽取,同时利用LDA(Latent Dirichlet Allocation)话题模型对科学文献的内容进行话题分析。此外,本文还提出了一个基于文本摘要和引用关系的可视辅助文档阅读系统。系统通过文本摘要技术提取论文中重要的句子,并利用多尺度的可视化方式展示出来,方便读者在阅读时定位到论文的核心内容;利用话题模型抽取出参考文献的核心话题,并设计多种可视化方案,包括词云,树图,径向图等,来展现参考文献的核心话题及其与该篇文章之间的关联关系;记录用户在整个阅读过程中的行为从而使用户关注在自己的阅读目的上,防止迷失。同时,我们在一个具体的使用场景下详细介绍了系统的使用方法以及交互方式,并进行了用户研究来验证系统的可用性,结果证明本文提出的系统具有可扩展性以及良好的用户体验。最后,在案例研究中,我们通过分析不同用户的阅读路径得出了许多不同的阅读模式,在未来的工作中将对用户阅读行为进行建模分析,并利用用户阅读数据做阅读推荐。
[Abstract]:In recent years, the number of scientific and technological papers has been increasing. Researchers need to read more and more literature. How to read a scientific paper quickly and effectively has gradually become an important research question. And a scientific paper is usually the crystallization of an academic research. It covers a lot of arguments and discoveries. This makes it difficult for readers to get the core arguments of an article in a very short time. On the other hand, while reading scientific papers, Understanding the important references that are relevant to them can be of great help in better understanding the content of the article. However, The citation network of an article is a complex structure. It is easy to get lost in the complex network of references to find the most important and relevant articles from the numerous references. How to not be lost in the multidimensional space of the document in the process of reading is also a problem worth studying. This paper makes use of the technology of text visual analysis to study the reading of scientific literature. In order to solve the above problems, Based on the study of the relationship and characteristics of scientific literature in the citation network and on the basis of the relevant research and techniques of document analysis, this paper proposes a text summarization technique based on reading purpose to extract the key sentences in the article. At the same time, we use the LDA(Latent Dirichlet allocation) topic model to analyze the content of scientific literature. This paper also proposes a visual assistant document reading system based on text summary and reference relationship. The system extracts important sentences from the paper by text summarization technology and displays them in a multi-scale visual way. It is convenient for readers to locate the core content of the paper while reading, extract the core topics of the reference document by using topic model, and design a variety of visualization schemes, including word cloud, tree map, radial map, etc. To show the core topic of the reference and its relationship with the article, to record the user's behavior throughout the reading process, so that the user can focus on his reading purpose and prevent him from getting lost. At the same time, In a specific usage scenario, we introduce the usage method and interaction method of the system in detail, and carry out user research to verify the usability of the system. The results show that the proposed system is scalable and has a good user experience. Finally, in the case study, we get a lot of different reading patterns by analyzing different users' reading paths. In the future work, the user's reading behavior will be modeled and analyzed, and the user's reading data will be used for reading recommendation.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.1

【相似文献】