Research And Implementation Of Text Matching Based On Multi-layer Attentive Interaction

With the rapid development of the industrial society,a large number of equipment failure case maintenance records have been accumulated in the daily maintenance work of power transmission and transformation equipment,including the knowledge of failure case discovery,judgment,and maintenance,which is quick for on-site operators when they encounter new failures.Judgment and maintenance decisions have important reference value.Text matching technology can dig out cases with high similarity to the target failure case from a large amount of failure case data,and provide key knowledge to on-site maintenance workers to assist in the decision-making process of failure maintenance.At present,text matching technology has made good research progress.Most models use semantic coding methods to extract text features and use text similarity calculation methods to predict the logical relationship between text pairs.However,there are insufficient semantic information extraction and uncaptured text pairs.The problem of mutual information between more detailed text and words.Aiming at these two problems,this paper proposes an efficient text matching model,which separately models text semantics and semantic interaction,extracts deep semantic information and interaction information,and combines the features of the two to construct text matching with enhanced semantics.model.First,this paper uses a multi-head self-attention network to extract the independent semantic information of the text;then,for the encoded text semantics,this paper designs a multi-head interactive attention network to extract the semantic matching and interactive information between text and words;finally,this paper designs The feature fusion method is used to enhance the semantic features of independent semantic and interactive information fusion group layer.The experimental results show that the model proposed in this paper has achieved leading performance effects on 7 data sets including STSb,QQP,AFQMC,WikiQA,QNLI,SNLI,and SciTail,which proves the efficiency of the model.This paper applies the proposed model to the field of power grids,designs and implements a prototype system for pushing similar fault cases to the power grid.A data set of similar fault cases in the power grid field was constructed,and the effectiveness of the model in the power grid similar case push was verified through experiments,and the system test proved the practicability of the system.