Risk Identification Method for Power Customer Service Complaints Based on Artificial Intelligence Model

The evaluation criteria for power customer service work are customer satisfaction evaluations, and improving customer evaluation scores is an important indicator of ensuring the quality of power supply services. Most customer demands that have not been satisfactorily resolved arise during the customer service process, and how to identify customer demands in customer service conversations is the key to solving this problem. The voice materials used in customer service are the basis for determining whether customer demands are met, but the efficiency of understanding large-scale voice materials through manual services is not high. The use of artificial intelligence language analysis models to batch mine customer demands in speech materials can significantly improve the efficiency of speech material analysis. Design a complaint phrase text matching model, using text matching algorithm to match the text in the speech material with the target text, and evaluate the speech material based on the matching coefficient to determine whether the customer meets the requirements of unresolved difficult demands. However, the text matching model calculates the customer’s complaint risk based on the similarity of text topics, and cannot accurately obtain evaluation results. On this basis, an attention model is introduced to superimpose the connection relationship and attention coefficient of keywords on the text matching results, and a comprehensive evaluation method is adopted to improve the prediction accuracy of customer complaint risk.