Design And Implementation Of Intelligent Outbound Call System Based On Text Classification

Traditional outbound call systems use the system to make outbound calls,use manual agents to communicate with customers,then complete outbound tasks.It consumes a lot of labor,and the work of customer service staff is tedious.With the rise of artificial intelligence technology,applying Natural Language Understanding to the customer service industry is not only in the interest of enterprises,but also a way to liberate productivity.This system mainly intelligently transforms the traditional outgoing call system,and uses the technology of Natural Language Understanding(NLU)to construct a virtual intelligent customer service to replace manual customer service.At the same time,the introduction of business management enables the system to have multi-service processing capabilities.The use of script configuration files and command script execution is equivalent to configuring dedicated intelligent customer service for multiple different services.This system optimizes the results of voice transcription and corrects errors in text obtained from voice transcription.For the error correction of speech transliteration text,the n-gram model(n-grarm)is used to model commonly used word combinations.And the text of the candidate set is recalled through the calculation of the longest common substring and editing distance of the pinyin sequence.Finally,use n-gram to sort the candidate set to get the most likely text sequence to achieve the effect of error correction.In order to achieve the goal of natural language understanding,this system implements a binary classification problem through TextCNN text classification of word vectors and kd-tree search matching.Based on this,a Bi-directional Long Short Term Memory Network(Bi-LSTM)and TextCNN are combined using Attention to perform Exploration of multi-classification problems.The paper expounds the system in the order of requirements analysis,system design,system implementation and system testing,and finally verifies the practicability of the system.