Research On Public Opinion Analysis Of Chemical Pollution Incidents Based On Sentiment Analysis

The occurrence of chemical pollution events is closely related to life and health and ecological environment.However,with the passage of time,the frequency of chemical pollution incidents such as exhaust gas leakage,wastewater leakage,and chemical plant explosion has gradually increased.These incidents not only pose a serious threat to people’s lives and living places,but also cause extremely serious damage to natural resources and the ecological environment.Due to the seriousness of these incidents,people are paying more and more attention to them and expressing their attitudes and opinions through social networking platforms,especially when the scenes of destruction of nature,buildings or lives are reported graphically and presented to the public,people will express their reactions and emotions more directly.In the face of the continuous chemical pollution incidents,capturing people’s attitudes and opinions through Internet big data can not only analyze the development of public opinion after the occurrence of chemical pollution incidents,but also help relevant departments to formulate relevant policies to avoid such incidents.This thesis studies the comment data of topics related to chemical pollution incidents under the Weibo platform to understand netizens’ feelings about such incidents.The following work was mainly accomplished:(1)The public opinion data set of chemical pollution events is constructed.The comment data of Weibo related to chemical pollution events have different text lengths,irregular wording,ignoring syntactic and grammatical requirements,and are characterized by no complete semantic structure and no obvious emotional features.The pre-processing of the original data is completed by removing redundant data and deleting symbolic expressions,and a commentary dataset for chemical pollution events is constructed.(2)The BERT-Bi LSTM sentiment analysis model is proposed to solve the problem of the lack of complete semantic structure future in the comment data.By fine-tuning the BERT model,the dynamic representation of word vectors of opinion text is generated in the downstream task,and the rich information of text is captured in the embedding layer to solve the problem of static representation of word vectors.The dynamic word vectors are used as the input of the Bi LSTM model to perform feature extraction and obtain local and global semantic features of the text,while highlighting the text sentiment polarity for sentiment classification.The effectiveness of the BERTBi LSTM model is verified through comparative experiments.(3)Due to the problems of insufficient data granularity and heterogeneous text information,a TEXT GCN sentiment analysis model was proposed.The potential information and related structures between documents and words,and between words are deeply mined,and 6399 items from the opinion dataset are re-extracted for more fine-grained segmentation.The TEXT GCN model is used to tri-classify their sentiment tendencies.The experimental results show that TEXT GCN does not require additional word embeddings and can effectively learn and predict the embedding representation of words in documents.(4)The public opinion visualization method for chemical pollution events is designed using Tableau technology.The method extracts the key information of Weibo comment data by using the sentiment label data trained by TEXT GCN model to present the trend of public sentiment changes on chemical pollution events.And it is applied to the chemical pollution event that is the focus of this paper for public opinion visualization and analysis.