Abstract:  Focusing on the issue of high cost and low efficiency associated with manual processing of customer complaints by operators, a quantitative research method based on TF-IDF (term frequency-inverse document frequency) algorithm is proposed, aiming to efficiently and accurately identify the reasons for customer complaints. Jieba, combined with the custom dictionary and the list of stopword is used to extract key words from complaint worksheets. The top six key words with the highest TF-IDF values in each issue are obtained, and a set of key words is output, thereby enhancing the accuracy and efficiency of keyword extraction. Furthermore, by comparing this method with the sole use of TF and the application of the TextRank algorithm, the importance of IDF and the differences in algorithmic principles are highlighted. Results indicate that issues related to optical modems, routers, and set-top boxes widely exist in complaints. In terms of these issues, this study provides operators with relevant suggestions for improving products and services, which have certain value to operators’ managing and solving problems.

Key words:  , complaint worksheet, reason for complaint, keyword extraction, TF-IDF