Composite question and answer method based on Bi-LSTM and Chinese knowledge graph

Abstract

According to the composite question answering method based on the Bi-LSTM and the Chinese knowledge graph, a semantic analysis technology is improved and introduced into a question answering system, a composite question sentence decomposition method based on deep learning is used, complex long and difficult sentences are rewritten into a plurality of simple sentences, the simple sentences are answered respectively, the processing capacity of the semantic analysis technology on composite questions is improved, and the processing efficiency of the semantic analysis technology on the Chinese knowledge graph is improved. The functions of analyzing the composite natural language question sentences and generating the composite natural language answers are achieved, the intelligent understanding ability and accuracy of the question answering system are improved, the process of processing the composite question sentences by the question answering system has interpretability, semantic information of the original composite question sentences is enriched, redundant information of the original composite question sentences is removed, and the user experience is improved. The problem that complex questions with complex sentence pattern structures and various themes are difficult to answer and the problem that answer information in composite answer generation is lost are solved.

Description

Technical field [0001] The present invention belongs to the field of computer artificial intelligence applications, and in particular to a composite question and answer method based on BI-LSTM and Chinese knowledge. Background technique [0002] Question Answering System, QA) establishes a communication bridge of natural language and machine language. As an important way for human-machine interaction, its essence is a high-level information search form that can help users extract relevant information from massive knowledge. With the high-speed development and extensive popularity of artificial intelligence, the Q & A system is widely used in legal recommendation, medical diagnosis, search engine, customer service. [0003] The traditional question and answer system can only answer the problem of the length of the sentence, the simple questions of the topic is accurate, and for more complex question cannot be effectively understood, the reason is that the semantic resolution techn...

AI Technical Summary

Problems solved by technology

However, this analysis method cannot meet the needs of understanding complex long and difficult sentences. There are three reasons: in longer questions, there are more key information, and key information will be missed when using the neural network model for entity extraction; The poor matching degree of triples in the knowledge map makes it impossible to search for entities and entity relationships in the knowledge graph; the current question parsing method cannot understand long and difficult sentences from the perspective of sentence structure, resulting in the inability to construct a targeted sex answer

Images