Power supply service customer appeal text classification method based on capsule network

Abstract

The invention discloses a power supply service customer appeal text classification method based on a capsule network. The method comprises the steps of: prprocessing power supply service customer appeal texts; generating a power supply service customer appeal text word vector, and solving the problem of one word with multiple meanings; acquiring a dependency relationship between the characters based on the encoding of the character vector and the dependency relationship; performing appeal text feature fusion based on the character and word vectors to obtain a fusion vector; adopting a self-attention mechanism to obtain an interdependence relationship between word vectors; aggregating appeal text capsules by adopting an EMRouting dynamic routing algorithm to obtain a semantic information relationship between words; and finally, calculating loss values of the positive class and the negative class of the appeal text by adopting an interval loss function. According to the method, the textimportant feature extraction capability is improved, and finally the accuracy of power supply demand text classification in the power customer service work order is improved.

Description

Technical field[0001]The present invention relates to the field of electric power technology, in particular to a text classification of power supply service customer appeals based on a capsule network.Background technique[0002]The quality of power supply service is an important indicator to measure the service level of power companies. With the rapid development of the Internet, big data, and social platforms, it is imperative to improve the quality of power supply services. As the service hotline of electric power enterprises, 95598 communicates and exchanges with electric customers through this hotline every day, thus forming a large amount of textual data of customer appeals. The best way to improve 95598 customer service capabilities is to clearly understand the customer service needs and problems expressed in these customer appeal texts, so as to provide accurate, proactive and differentiated power supply services, improve customer satisfaction with electricity, and thus improv...

AI Technical Summary

Problems solved by technology

[0005] For CNN and LSTM networks in text classification, the feature expression ability is not strong, and they cannot break through the limitation of text distance to obtain the semantic and word order information defects of the context, and the power supply service customers appeal to the characteristics of unstructured text data, large amount of i

Images