瓴羊One客服会话质检分析数据
收藏资源简介：
对合法采集的企业服务会话数据进行分析、处理，帮助企业识别会话内容的质量与规范性，提升服务品质。瓴羊One客服会话质检分析数据的算法规则包括： 1、数据采集和处理：企业客户自身积累的客服和用户之间的语聊会话数据，对用户的id等敏感信息做匿名化处理。 2、算法加工：基于外部数据，涉黄、涉政等字典库，欢迎语等样本数据，训练用于识别涉黄、涉政、是否欢迎语等类别的NLP多分类模型。针对客服在服务用户过程中的语聊数据，基于辱骂、涉黄、涉政检测，是否使用欢迎语等NLP模型，识别客服服务过程中是否规范，服务是否有温度。 3、应用场景：客服服务质量需要进行考核，通过会话质检，一方面，可以检测客服在服务过程是否有辱骂用户、涉黄、涉政等高风险话语，另外，可以在服务过程是否有欢迎语等体现服务温度方面进行检测。
This dataset focuses on analyzing and processing legally collected enterprise service session data to help enterprises assess the quality and standardization of session content and improve service quality. The algorithm rules for Lingyang One Customer Service Session Quality Inspection and Analysis Data are as follows: 1. Data Collection and Processing: Utilize the voice chat session data between customer service representatives and users accumulated by enterprise clients themselves, and anonymize sensitive information such as user IDs. 2. Algorithm Development: Based on external resources including dictionaries for detecting pornographic and politically sensitive content, as well as sample data such as welcome messages, train an NLP multi-classification model to identify categories such as pornographic content, politically sensitive content, and whether a welcome message is used. For voice chat data generated during customer service-user interactions, apply NLP models for detecting abusive language, pornographic content, politically sensitive content, and the use of welcome messages to evaluate whether the customer service's service process is standardized and whether the service demonstrates warmth. 3. Application Scenarios: For enterprises that need to assess customer service quality, session quality inspection has two main applications: on one hand, it can detect high-risk utterances such as abuse, pornographic content, and politically sensitive content from customer service representatives during the service process; on the other hand, it can also check whether the service includes welcome messages and other details that reflect service warmth.
