Text similarity detection method of power customer service work order based on TFIDF algorithm
Du Lifeng, Changhua Hu · 2022 IEEE 5th International Conference on Information Systems and Computer Aided Education (ICISCAE) · 2022
The increase in the number of customer complaints has led to an increase in the number of power customer service work orders, resulting in the failure to solve the reasonable demands of some customers on time, affecting the customer service quality of Power Supply Bureau. To solve the above problems, a text similarity detection method of power customer service work order based on TFIDF algorithm is proposed. This method uses web scraper technology to obtain the power customer service work order text, and preprocesses the text, including Chinese word segmentation, desktop words and text representation. TFIDF algorithm is used to extract the preprocessed text keywords and obtain the keyword weight. Based on the weight, cosine distance method is used to calculate the text similarity of text keywords of different power customer service work orders. The results show that compared with the detection method based on weighted word and sentence vector, the detection method based on hybrid model, the detection method based on coupling distance discrimination and strong category features, the macro average F1 measurement value is higher, the method takes less time and has lower complexity.