Abnormal personality detecting method and device based on dialogues

Abstract

The invention provides an abnormal personality detecting method and device based on dialogues. The method comprises the steps of obtaining a first preset number of pieces of dialogue data in social media; utilizing a support vector machine (SVM) for performing emotion recognition on the first preset number of pieces of dialogue data, and obtaining the public emotion and each user's emotion under asecond preset number of emotions; marking the public emotion under the second preset number of emotions and each user's emotion; on the basis of marks of the public emotion and each user's emotion under the second preset number of emotions, determining emotion transfer tensors of the public and each user; performing probability statistics on the emotion transfer of the second preset number of dialogues, and obtaining the emotion transfer tensors of the public and each user; according to the emotion transfer tensors of the public and each user, calculating the tensor similarity; according to tensor similarity and a similarity threshold value, determining abnormal user individuals. It can be seen that the abnormal user individuals can be rapidly determined.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a dialog-based abnormal personality detection method and device. Background technique [0002] The historical dialog data in related technologies mainly includes the following schemes: [0003] Firstly, by collecting and labeling historical dialogue data, the anomaly detection model is trained using the labeled data, and when real-time dialogue data is received, the trained anomaly detection model is used to perform anomaly detection and obtain results; the training of the anomaly detection model is Refers to: By reducing the anomaly detection problem to a binary classification problem (I means there is anomaly, O means there is no anomaly), and use a machine learning model, such as a support vector machine (SVM) or neural network to train a two-classifier, that is, an anomaly detection model . However, the above method ignores factors such as the difference between eac...

AI Technical Summary

Problems solved by technology

However, the above method ignores factors such as the difference of each user and the difference of external stimuli, and only trains the model based on historical conversational data, resulting in inaccurate detection of abnormalities.

Images