Research On Natural Question And Answer Model For Civil Aviation Airport Passengers

The airports are at the forefront of technological innovation,because the number of passengers traveling by air is increasing exponentially each year mainly.At present,digital information technology has brought great changes to social development,"smart city","smart transportation" and other concepts have gradually changed from theory to reality.In this change,the civil aviation industry also complies with the requirements of the times and implements the transformation of information and intelligence.Among them,passenger-oriented service is the most important.Intelligent question and answer system can help passengers get information quickly,reduce the cost of information acquisition,and improve passenger satisfaction.Therefore,it is of great significance to study an efficient question and answer model.There are two main parts about the construction of civil aviation question and answer model: named entity identification and relationship extraction.A method for civil aviation entity recognition combining Bi-directional Long Short-Term Memory(Bi LSTM)and Conditional Random Field(CRF)is designed.The real civil aviation question and answer dataset obtained by the crawler is labeled entirely,and the labeled data is transformed into word vector.Using Bi LSTM model to get the contextual information of civil aviation text;Finally,the results are obtained from the CRF model.The results of comparative experiments show that this method has a good effect in the field of civil aviation,and effectively improves the F1 value.On this basis,a method of extracting text entity relationship under self attention mechanism is proposed.All words are mapped to low-degree real vectors through a text vector model,so that the text can be transformed into vector mode,and all words in the sentence can be transformed into embedding matrix by learning embedding according to the external situation of the words.The Bi LSTM network is used to create text vectors,access the previous and future contexts,and fuse the two output vectors.The activation function is used to compress the word dimension,calculate the semantic contribution weight of the upper and lower text of the sentence,add the self attention mechanism between the Bi LSTM layer and the output layer,obtain the sentence semantics of the matrix level from multiple angles,calculate the score of the feature vector of the combined sentence on the relationship,and complete the extraction of the text entity relationship according to the given probability random sampling weight parameter variable.Experimental results show that the proposed method has a good effect on the extraction of text entity relations and has high accuracy.Finally,a natural question-and-answer model for civil aviation airport passengers is designed and constructed.Using the prior knowledge of Knowledge Embeddings(KEs),a natural question and answer model for civil aviation airport passengers is built.By building the set of candidate entities and the set of candidate relationships,using the attention mechanism,combining the information of the question with the candidate entities and the candidate relationship,the entity-relationship pair with the highest score is predicted,and the final answer entity is found.By comparing with other models,the knowledge embedding model based on bilinear model performs better.Experiments show that the model performs better than the traditional model.