Research And Application Of Water Resources Information Intelligent Question Answering Based On Knowledge Graph

With the development of information technology,my country’s water resources industry is constantly moving towards modernization,and a large amount of multi-source and heterogeneous information has been accumulated in the field of water resources.Information required.The traditional information acquisition method is mainly based on search engines,but the answers often return document collections.For users,they need to read the document content before extracting the answer.In recent years,knowledge graphs have attracted high attention in academia and industry because of their strong expressive ability and good expansibility.Knowledge graphs provide high-quality knowledge bases for intelligent question answering systems and promote the development of question answering systems.Knowledge graphs are widely used in many fields,but relatively few in the field of water resources.In this paper,by constructing a knowledge graph in the field of water resources and applying it to the question answering system,the intelligent retrieval service level of water information information is improved.Its research contents are as follows:1.Build a water resources knowledge map.Aiming at the Chinese corpus in the water resources field,this paper constructs a named entity recognition method in the water resources field based on BERT-Bi LSTM-CRF.As an end-to-end deep learning pre-training model,the BERT model can extract rich text features to obtain output vectors,and then use Bi LSTM to capture the semantics of each word in the context,and finally use the CRF model to extract the optimal sequence.This paper proposes an entity relation extraction method based on co-occurrence network for unstructured texts in the field of water resources.Since the construction of relation extraction based on supervised learning relies on manually annotated data,the unsupervised learning method in this paper does not require any manual annotation.Suitable for the ever-growing web text of the Internet.The relevant experiments show that the method proposed in this paper has a good extraction effect for information extraction of water conservancy unstructured text.2.Design and implement a question answering algorithm based on knowledge graph.The question answering algorithm is decomposed into two tasks of named entity recognition and question answer classification in the field of water resources.BERT-Bi LSTM-CRF is used to identify entities in the question sentence,paving the way for subsequent question classification.Combined with question search engines such as "Baidu Knows",the characteristics of questions are analyzed,domain questions are generated,and the performance of various classifiers is compared and analyzed.Through analysis and comparison,it is found that the convolutional neural network classifier has the best classification effect as an intent classification model.Finally,the question classification results are combined with the Neo4 j database,and the answer search is completed by querying the knowledge base through Cypher statements.3.Design and implement an interactive water resources knowledge graph question and answer system.Using Django as the web service framework,the question and answer algorithm is connected to the We Chat public account to automatically reply,and answers are given to the questions raised by users in the field of water resources.In this paper,by constructing a knowledge graph in the field of water resources,combined with the deep learning method to realize knowledge question and answer,a question and answer system based on the knowledge graph of water resources is realized,which can provide users or decision makers with fast and accurate information retrieval services,and improve the level of water information management and information services.