Research On Scientific Monitoring Based On Topic Model

In the era of knowledge economy and big data,with the rapid development of Sci-Tech and research results, how to the provide effective scientific monitoring services from the mass scientific literature has received great attention, so as to provide effective Sci-Tech support for research management departments and academic researchers.On one hand, quantity of scientific literature grows exponentially,becoming a huge storehouse;on the other hand, faced with massive and fast-growing scientific literature data the needs to extract knowledge and information from the literature databasesquickly and efficiently becomes very urgentHow to conduct scientific monitoring from massive information environment to seize scientific developmenteffectively, grasp the coreand the research focus timely is a key issue.At present, research on scientific monitoring uses scientific journals and patent literature as main research object, applies bibliometric and social network analysis as main approach. To make data sources more diverse, research methodsmore sophisticated, research findings more explanatory power for research on scientific monitoring, you need to apply a variety of scientific literature carriers and continue to break the existing research methods.Therefore, based on the systematic investigation and comprehensive analysis of the existing research,this paper proposes effectiveand scientific literature orientedscientific monitoring methodsbased on LDA topic model.Mining the scientific knowledge from scientific literature,in order to reveal the evolution of scientific topics and context purposes, grasp scientific developments to enhance the value of its Sci-Tech innovation.Firstly,the paper starts with the "Topic",then uses scientific reporter as the example object to conduct topic analysis and model the three layer topology "Document-Topic-Word ".Secondly,study how to automatically mining "Topic" from massive scientific literature,the time sequence ofevolutionand other scientific monitoring related issues,including topic mining,topic strength measurement,topic number determination and topic evolution. Thirdly, design and implement prototype system based on the open source platform JGibbLDA.Finally, based on the topic model related research, take the similarity computing as a breakthrough, document similarity calculation model and recommended model based on user interest are proposed to improve derived application oriented scientific monitoring.