Research On The Key Technologies Of Question Answering Over Domain Knowledge Graph

Recently,the rapid development of knowledge graph enables data to be organized and understood in a form that is closer to human cognition,providing high-quality data resources for the application of artificial intelligence.As a typical application of knowledge graph,question answering uses the entities and their relations to reason and get the answer.It can accept the questions described in natural language form,better understand the real intention of questions,more effectively meet their accurate information needs,and provide personalized knowledge services for them.Taking Chinese poetry as an example,mainly study the key technologies of question answering on Chinese poetry knowledge graph: entity linking and entity relation extraction.The specific research contents are as follows:(1)Multi-feature fusion entity linking based neural network.According to the disadvantages caused by word segmentation and the inability to make use of the word information in a sentence,adopting the Lattice-LSTM model.In view of the deficiency of entity reference features,integrates popularity features,entity context features,entity type features,name similarity features and inclusion character features,and selects the entity with the highest score as the target entity.The experiment demonstrates that the method is better than some traditional methods in accuracy.(2)BIGRU entity relation extraction based double attention.In view of the shortcomings of original features,enriching text features by integrating word features,part-of-speech features and position features in sentences in the feature extraction stage.Secondly,introducing the attention mechanism based on words and sentence into the traditional Bi GRU network,so as to make the model pay more attention to the more important features and reduce the influence of noise on the extraction task.Experimental results show that the model can improve the accuracy of entity relation extraction.(3)Construct Poetry knowledge graph.For the lack of Poetry knowledge graph,first analyz the service demand of Poetry knowledge graph,then design the poetry schema,second construct poetry knowledge graph through data collection,knowledge extraction,knowledge fusion,knowledge storage and other technologies to provide data support for the poetry knowledge question answering.(4)Based on the above research,the poetry knowledge question answering based on the semantic parsing is built,and realize the visualization of the knowledge graph.