Research On Improved Aspect-level Opinion Mining Algorithm And Visual Analysis Based On Dependency Relationship

The development of the Internet has brought a lot of comment text data,but it is often very difficult to quickly and accurately find valuable information in a large amount of unstructured data.Opinion mining can help users extract valuable information from a large amount of text data.However,most of the current researches tend to categorize praise and criticism as a whole sentence,lacking fine-grained aspect-level opinion mining and efficient summarization methods.This paper conducts aspect-level opinion mining on mobile product review data of an e-commerce platform.The main research content includes the following aspects:First,in view of the aspect extraction problem in aspect-level opinion mining,Py Corrector is used to correct the error of the comment text and improve the accuracy of the mining results.Through lexical analysis,reliable frequent aspects are selected as the seed word set.The word embedding model is incrementally trained using mobile phone corpus,the infrequent aspects are extracted using the vector similarity measure,and the aspects are clustered through the K-means algorithm,which enhances the comprehensiveness of aspect extraction and classifies the aspects.Secondly,in view of the fact that there are few researches on aspect-level opinion mining and lack of specific viewpoint content,an improved aspect-level opinion mining algorithm based on dependency relationship is proposed.Incorporate lexical analysis and aspect extraction results on the basis of dependency analysis to improve the accuracy of mining.By analyzing the dependency relationship and summarizing the extraction rules,the effective extraction of aspects and viewpoint content has been completed.Thirdly,in view of the lack of efficient viewpoint summarization,visual design is carried out,and complex relationships are effectively displayed by constructing a knowledge graph.Combine the mining results and user needs and goals to design a visual analysis system to efficiently summarize and display product information and comment mining results from multiple angles.Finally,design experiments to evaluate the mining effect of the algorithm,verify the effectiveness of the proposed method,and analyze the differences and reasons in the evaluation results.Then combined with practical applications,the product review mining results visualization system is analyzed,which verifies the practicability of the aspect-level opinion mining and visualization system.