Research On Topic Tracking Algorithm Based On "Feature Reduction Dimension" Text Complex Network

With the continuous iteration of mobile Internet technology,the mobile Internet business has also exploded,which has a huge impact on people's daily lives.In the field of mobile social networking,applications represented by Weibo,WeChat,and Today's headlines play an important role in connecting and enabling efficient linkage between people and people.Accurate content push is also an important guarantee for enhancing the user's stickiness to the platform.Users get information from all parties at a very low cost and publicly express their opinions through various media platforms.Over time,users enjoy the convenience and endure the troubles caused by the complexity of information.In the context of big data,how to quickly and accurately capture the social hotspot information of concern has become the core issue of network public opinion development.Therefore,topic tracking technology emerges as the times require,which aims to achieve dynamic tracking of known topics in the news media information flow,and penetrates into various public opinion analysis systems.This paper starts from the current practical problems,clarifies the research background of the subject,solves the main problems,and then clarifies the research significance.This paper proposes a topic representation model of "feature reduction dimension" text complex network,which solves the problem of high dimensionality and sparsity of traditional word co-occurrence complex network processing large-scale corpus,and proposes a topic for network.In the research of topic tracking algorithm based on text complex network,considering the structural features and statistical features of semantic network,the research on the weight calculation method of network model nodes is carried out to solve the problem of inaccurate weighting caused by the unilateral contribution of traditional algorithms to one-sided investigation of words.Specific innovations:(1)Propose a feature dimension reduction method combining Chinese semantics and grammar rules,innovate the complex network generation rules of texts,and improve the application effect of complex network models in text representation problems.(2)Simulating the deduplication principle of web pages,introducing the simhash similarity calculation algorithm.Designed to reduce the computational complexity between multiple networks,thereby improving the accuracy of topic tracking.