Dialogue pre-labeling method and system, computer equipment and storage medium

The invention provides a dialogue pre-labeling method. The dialogue pre-labeling method comprises the steps of obtaining dialogue data; the dialogue data is obtained based on voice recognition; carrying out preprocessing on the dialogue data; inputting the preprocessed dialogue data into a pre-labeling model, and pre-labeling the dialogue data; wherein the pre-annotation model is obtained based on prompt learning and training of a pre-training language model. According to the scheme, for the data set without labels, the pre-training large model is used before the labeling stage, zero sample pre-labeling is carried out in a Prompt mode, the overall business execution efficiency is improved, the absolute influence of manpower in the business execution process is faded, and the data quality is indirectly improved. 本发明提供了一种对话预标注方法，包括：获取对话数据；所述对话数据基于语音识别得到；对所述对话数据进行预处理；将预处理后的所述对话数据输入预标注模型，对所述对话数据进行预标注；其中，所述预标注模型基于提示学习和预训练语言模型训练得到。本方案针对没有标签的数据集，在标注阶段之前使用预训练大模型，通过提示(Prompt)的方式进行零样本预标注，提高业务整体执行效率，并淡化人力在业务执行过程中的绝对影响力，