Research And Application Of Intelligent Question Answering In The Government Affairs Fiel

With the rapid popularization of technologies such as big data and artificial intelligence,along with the accelerated digitalization of government services,an increasing number of administrative services are transitioning from traditional offline processing to online processing.This shift enables data to travel more while reducing the need for citizens to physically visit government offices.However,as the number of online administrative tasks increases,the volume of messages on government service websites has also surged.To alleviate the pressure on human operators,government service websites in various provinces have started implementing government Q&A services.Existing government Q&A services primarily rely on traditional keyword matching techniques and building databases of frequently asked questions.Keyword matching retrieves corresponding entries from the database based on the keywords,offering broad coverage but lacking in providing fine-grained answers.On the other hand,building a database of frequently asked questions can provide detailed answers to common queries but has limited coverage.To address these issues,this study proposes intelligent question-answering methods based on domain knowledge graphs and deep learning techniques in the field of government services.These methods leverage government knowledge graphs for intelligent question answering and employ machine reading comprehension techniques for government Q&A.This approach ensures broad coverage of questions while also obtaining fine-grained answers,thus enhancing the effectiveness and intelligence level of government Q&A.To apply these two intelligent question-answering methods in the domain of government services,the following research work was conducted in this study:(1)Intelligent Question-Answering Method based on Government Knowledge Graphs.The data of administrative tasks in government service platforms is semistructured.In this study,by defining the structure of government ontology,it can be directly transformed into structured knowledge graph data.As there is no need for entity recognition and relation extraction steps,it reduces error accumulation and ensures the quality of the knowledge graph.Subsequently,based on the government knowledge graph,this study proposes a retrieval ranking-based knowledge graph question-answering method,which mainly consists of entity mention recognition and question intent identification.Entity mention recognition aims to accurately identify entity mentions for better linking to relevant entities in the knowledge base.In this study,the Mac BERT-Bi LSTM-CRF model is designed to achieve entity mention recognition.Question intent identification aims to determine the intent of the question and provide the final answer.To better differentiate the differences between intents,this study utilizes the Sentence-BERT model based on pre-training and siamese networks for question intent identification.(2)Government Intelligent Question-Answering Method based on Machine Reading Comprehension.The procurement notices and bidding announcements issued by government departments are unstructured data.To address this type of data,this study designs a machine reading comprehension method based on the Chinese BERTBi LSTM model to achieve government question-answering.Additionally,there are variations in the writing styles of procurement notices and bidding announcements across different government departments.Some of these notices and announcements may lack standardized language.Therefore,data cleansing techniques were applied to address this issue.Furthermore,considering the limited availability of manually annotated data,a self-training data augmentation method was utilized to expand the dataset.(3)Design and Implementation of Government Intelligent Question-Answering System.This paper conducts an in-depth analysis of the requirements for government question-answering and builds a question-answering system in the government domain by combining the intelligent question-answering method based on government knowledge graphs and the government question-answering method based on machine reading comprehension.During the implementation process,comprehensive testing was conducted to evaluate both the functional and nonfunctional aspects of the system.The test results demonstrate that the developed system effectively meets the design requirements.By incorporating intelligent question-answering techniques into the government question-answering system,the system facilitates the intelligent development of government question-answering and holds significant importance in achieving efficient public services.