Design And Implementation Of Ship Berthing And Unberthing Knowledge Question Answering System Based On Natural Language Understanding

In the process of the continuous development of artificial intelligence technology,question answering system has always been one of the hot research directions of researchers because of its wide application scenarios.In the era of rapid growth of information,the cost of obtaining information on traditional search engines is increasing.Specifically,the answers returned by search engines are not intuitive and accurate,but mostly related pages matching keywords.Users need to spend more time filtering to find the answers.The query results of the question and answer system are precise and concise,and this form is more in line with the needs of users.Knowledge graph is a collection of human knowledge.Compared with traditional knowledge storage forms,Knowledge graph is composed of nodes and edges,which constitutes a huge relationship network in semantics,and expresses things in the real world and the relationship between things in a way more in line with human thinking.The Q&A system based on the Knowledge graphs takes into account the advantages of Q&A system and Knowledge graphs,providing more possibilities for improving the user experience.Ship berthing and unberthing is an important stage in maritime transportation task.Aiming at the knowledge needed before berthing,this paper designs and implements a ship berthing and unberthing Knowledge Q&A system based on Knowledge graph,and focuses on the natural language understanding part of the Q&A system.Firstly,in order to build a Knowledge graph in the field of ship berthing and unberthing knowledge,this paper uses the crawler framework to collect the original data in port,ship and other websites,and completes the task of knowledge extraction through entity recognition,relationship extraction,attribute extraction and other technologies.Then,the constructed triplet data is imported into the neo4j graph database to complete the construction of the Knowledge graph.Secondly,the key technologies of natural language understanding module in question answering system are studied.In the understanding task of ship berthing and disembarking in the single-round dialogue scene,it can be divided into two subtasks:entity recognition and question classification.The Bi-LSTM-CRF model used for entity recognition.In view of the poor effect of the question classification baseline model in the scene with unbalanced data samples,from the perspective of improving the feature extraction ability of the model,In the feature representation of words,cyclic neural network is used to embed contextual semantic information.At the same time,in order to improve the ability to capture the semantic information of longer sentences in question classification,the Transformer algorithm,which introduces attention mechanism,has been significantly improved compared with the benchmark model in the task of text classification.Then,the dialogue understanding task in the multi-round dialogue scene of berthing and disembarking is studied.It is divided into user intention recognition and semantic slot filling annotation.Aiming at the problems of insufficient utilization of information features and loss of long-distance semantic relationship in the baseline model,intention attention and full mode attention mechanism are integrated into the baseline model to improve the ability of semantic capturing information.In addition,the joint training model integrated with the pre training mechanism is used.The feature representation ability of Bert model is better than that of ordinary word vector.At the same time,the deep two-way Transformer component has more advantages for capturing the dialogue context information.Experimental results show that the improved algorithm has achieved good results in both Chinese and English data.Finally,the algorithm model is applied to the ship berthing and unberthing Knowledge Q&A system,which can complete the tasks of information Knowledge graph visualization,question query and system Q&A.According to the different functions of the question and answer system,several plates of the human-computer interaction interface are designed,and the multi-form interaction mode of voice input and manual input is provided to facilitate the use of users.