| 摘要 |
The invention discloses a man-machine conversation satisfaction degree prediction method and system. The method comprises the steps that satisfaction degree training data of a plurality of training man-machine conversations are obtained; conversation characteristics of the training man-machine conversations are adopted as training samples, corresponding satisfaction degree results are adopted as target values for classification model training, and a classification model is obtained; an actual man-machine conversation is obtained, the conversation characteristics of the actual man-machine conversation are extracted, and the conversation characteristics of the actual man-machine conversation are input into the classification model to obtain the satisfaction degree result predicted by the classification model. When the satisfaction degree is predicted, the conversation characteristics influencing prediction can be added or deleted easily, through a statistical mode, the weights, needing to be specifically set, of all the characteristics do not need to be directly considered any more, only the proper classification model needs to be selected, and weight setting is automatically completed through the study of the classification model on the training samples. |
