Research And Implementation Of Multimodal Network Public Opinion Abstractive Automatic Summarization Technology

In the era of big data with the rapid development of information technology,massive and rich multi-modal information(such as images)has gradually replaced the traditional single text modal information,which is the object that public opinion supervision departments need to focus on.It is of great significance to analyze and extract key information of multimode network public opinion to help public opinion workers quickly grasp the key content of massive information.Aiming at the problem that the complexity of multi-modal online public opinions is not conducive to the centralized expression of subject content,a multi-modal feature extraction model of online public opinions is designed and implemented.The feature extraction technology is used to convert different modal information into a single text modal information containing key information,so as to obtain more comprehensive original information for subsequent generation of abstract.The experimental data and results show that the BLEU value is increased by 0.2395 on average,and the analysis of the extraction results also shows that the results of the model are suitable for the work requirements.Aiming at the problem of scattered semantic expression of multimode online public opinion generative summary,a Chinese generative summary model integrating the attention mechanism of domain core words was designed and implemented.By increasing the attention to the domain core words,the understanding and expression of key information were enhanced,so as to obtain the public opinion summary with high accuracy and strong domain core.The experimental data and results show that the index values of ROUGE-1,ROUGE-2 and ROUGE-L generated by the model are respectively increased by 0.0442,0.0151 and 0.0043 on average compared with other comparative abstract models.The content of the indicates that the results are more accurate and more suitable for public opinion work after analysis.Abstract comprehensive evaluation method is improved and applied to solve the problem that automatic summary general evaluation method is not suitable for professional evaluation of public opinion.By designing and realizing the professional evaluation index of the abstract,the professional score of the abstract can be obtained,and a more comprehensive evaluation of the abstract can be carried out to make the abstract more suitable for application in the field of public opinion.The experimental data and results show that the average score of the index on the standard abstract set is 94.27,and the score fluctuates within±5,which can reflect the core and stability of the content of the abstract to a certain extent,and is not susceptible to extreme circumstances,which is conducive to the application of public opinion work.Finally,the multi-mode online public opinion generation automatic summary technology is applied to the online public opinion monitoring and early warning system,which can meet the needs of public opinion after testing.Through automatic summary,public opinion staff can quickly obtain the core content of massive multi-modal information,timely grasp key information,discover problems and respond to them,improve the efficiency of public opinion work,and make positive contributions to promoting the health and stability of the network environment.