Font Size: a A A

Application Of Telecom Work Order Association Based On Short Text Similarity Recognition

Posted on:2023-07-15Degree:MasterType:Thesis
Country:ChinaCandidate:J M FengFull Text:PDF
GTID:2568306794487584Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
As a typical task object in natural language reasoning,text similarity judgment has a very wide range of scenarios in practical applications.In recent years,with the product of comprehensive services of the three major telecom operators,the number of services is increasing.There are different groups of frequent alarm problems,but multiple similar alarms are derived from a piece of single alarm information and the same alarm occurs repeatedly.It makes manual processing difficult.Aiming at the urgent need of the Inner Mongolia Telecom Branch to solve the weak correlation between reducing invalid alarm work orders and traditional alarm information correlation methods,this paper applies text similarity calculation to the work order similarity correlation of the Telecom alarm system.Based on the ESIM(enhanced LSTM for natural language influence)model,a transformer is used as the feature extractor to replace the feature extraction part of the ESIM model.According to the combination of the transformer’s self-attention mechanism and ESIM text interaction characteristics,alarm work order correlation is realized.Compared with the ESIM model,the accuracy of alarm work order similarity calculation is improved by 1.7%.Secondly,the similarity of the Telecom alarm work order is related to multiple characteristics of the work order.Based on the analysis of the data characteristics of alarm work order,this paper proposes a calculation method of alarm work order similarity integrating multiple characteristics.It can fully extract the amount of alarm work order information and improve the processing speed of the model under the condition of ensuring high accuracy of similarity association.Integrate the alarm description,network element name,and first alarm time characteristics,give different weights according to the importance of different characteristics,and realize the work order similarity calculation after normalization,to make the model more in line with the actual needs of telecommunications.Finally,the work order matching system is constructed.Login module,alarm matching module,fault cause feedback template module.Realize the similarity association of a given alarm work order at the same time,and provide users with the causes of alarm faults.
Keywords/Search Tags:Text similarity, ESIM, Transformer, Multi-feature recognition, Alarm work order
PDF Full Text Request
Related items