Research On Method Of Constructing Temporal Topic Chains Based On The Scientific Literature

Scientific literature is one of the most important forms for showing scientific and technology research results. And it is a very significant carrier for transferring information and academic idea communication. Researchers often acquire knowledge, carry on research and make scientific decisions through reading related literature. Scientific literature is the foundation of scientific research for researchers. As a concise summary for content of the literature, a topic can represent and explain what the literature talk about. As time goes, topics always evolve in the literature gradually. During the process, new topics may emerge, old topics may disappear and some topics may transfer or converge to other topics. Showing topic evolution process in the literature can help researchers to understand rules or routes for the scientific study and make scientific decisions.In the thesis, a temporal topic chain is used to reveal the fact that topics often change over time. We analyze topic evolution process and propose a method of constructing the temporal topic chain. An empirical study is conducted based on scientific literature. Besides, we will analyze characteristics of intersection or combination between topics qualitatively and quantitatively.Firstly, the model of temporal topic chain is introduced and its corresponding construction method from a text document with time stamps is proposed. The temporal topic chain is expressed as a weighted directed graph. Considering the fact that a new topic can be more easily and fast transmitted, a new method for calculating the defined transition probability between topics is proposed. In the method, an attenuation factor is introduced and it is measured by Ebbinghaus Forgetting Curve. To completely reveal routes for topic evolution, we make some rules and propose a new temporal topic chain construction method.Secondly, the empirical study has been conducted on the scientific literature in a specific field. The results shows that a temporal topic chain can effectively expose how a topic change over time and also reveal some latent relationships between topics during topic evolution process.Thirdly, the thesis analyzes characteristics of intersection or combination between topics existing in the temporal topic chains on the empirical study. And a quantitative method for the extent that some topics intersect or combine with each other is proposed. There are some integrated structures in a temporal topic chain and this kind of structure reveals characteristics of intersection or combination between topics.A temporal topic chain can expose rules or routes for the scientific study and implicit relations between topics. Besides, analysis about characteristics of intersection or combination between topics helps researchers to discover innovative topics and new ideas for solving problems. It is important for researchers to acquire this kind of guidance and inspiration.