Research On The Key Technologies Of Military Scenario Event Extraction

It is an urgent problem in the field of military information processing to quickly and accurately obtain the required effective data information from a large number of military texts and transform it into a structured form that people and machines can understand and use together.Information extraction is an effective method to solve this problem.Event extraction is the key to realize information extraction.Military scenario event extraction refers to the process of extracting event information required by users from military scenario texts expressed in natural languages and presenting it in a structured form.It is the application of information extraction,event extraction,data identification and other technologies in the field of military scenario.It can help military personnel to accurately and efficiently understand and handle military scenarios and provide support for military scenario application and operational training data construction.This paper studies the key technology of military scenario event extraction which provides a technical approach for military scenario event extraction and lays a foundation for related data identification and information extraction system development.1.The basic problem of military scenario extraction is studied.According to the characteristics of military scenario,the connotation and basic concepts of military scenario event extraction are defined.11 types of military scenario events are defined by list enumeration method and structural tree diagram analysis method.The overall framework and basic process of military scenario event extraction are given,which lays a good foundation for military scenario event extraction.2.An iterative construction method of military scenario event extraction corpus based on machine prediction and manual tuning is proposed.In order to solve the problem of lack of mature corpus in this field,combining the advantages of automatic corpus construction and manual corpus construction,the corpus is constructed through continuous iteration of machine prediction and manual tuning.The experimental results show that the method is feasible and effective,and improves the efficiency of corpus construction.3.Design and construct a military scenario event extraction model and algorithm based on N decoding.Using the second-order hidden Markov model and the improved Viterbi algorithm,a military scenario event extraction model and algorithm based on onetime decoding constructed.Aiming at the problem that the matching degree between event trigger words and event elements is not high,the model and algorithm based on once decoding are improved,and the military scenario event extraction model and algorithm based on secondary decoding are proposed and constructed.To solve the problem of sparse data,a data smoothing method is adopted.Experiments show that the model and algorithm are effective in solving the problem of military scenario event extraction.4.Construct a military scenario event extraction model and algorithm based on BGRU-Attention.Aiming at the problems of low flexibility and sparse data in traditional event extraction methods based on machine learning,a military scenario event extraction model and algorithm based on BGRU-Attention is constructed by using bidirectional GRU neural network and fusing double-layer ATTENTION mechanism.The model and algorithm can mine deep-level semantic feature information and obtain good experimental results.