Research On Content Linking Based On Multi-feature Fusion And Deep Learning

With the rapid growth of Internet information,user requirements for information retrieval are also getting higher and higher.We need toeliminate redundant information and obtain accurate information,avoiding a great deal of time and effort spent in filtering out and sorting out useless information or known information.The content linking research is just to solve the above problems.It aims to connect the paragraphs or sentences together in different texts to make people understand the specific texts more comprehensively,no longer need to find relevant content manually and refine related information.This thesis uses the corpus of scientific papers and social network contents to study content linking.Papers mainly refer to scientific research,and social network contents tend to be built for commercial applications.For the above two aspects,content linking has good practical value and development prospects.For the study of content linking,this thesis explores both the syntactic information and semantic information.At present,most content linking research is based on syntactic information.Although it is superficial information in natural language,it has relatively good performance.Semantic information is the future trend of research,and we have realized the importance of in-depth semantic level.Therefore,this thesis studiescontent linking from two aspects of syntactics and semantics.For semantic aspect,we attempt to combine word vector with convolutional neural network,while for syntactic aspect,we fuse many kinds of text similarities together.Experiments have shown that the proposed methods haverelatively good performance and certain research significance.