Personality Detection Technology And Research Based On Affective Knowledge Enhancement

Personality is a stable psychological structure that reflects individual’s choices and preferences and affects all aspects of an individual’s life.Personality detection refers to predicting someone’s personality type by analyzing their behaviors and preferences,which has been widely used in recommendation systems,social network analysis,and personalized products and services.Therefore,the automatic detection of personality has gradually become an important subtopic in the field of artificial intelligence,and has also become an important concern in the field of social engineering in cyberspace security.There are two main limitations in the existing personality detection models for text data.On the one hand,there are few text datasets available for personality detection,and it is difficult to extract the deep associations and features of language structure only through traditional features,and the external large-scale unlabeled information is not fully utilized.On the other hand,most of the existing research focuses on using deep neural networks to extract the semantic representation of text,ignoring the attention to emotion and psychological characteristics,and not making full use of the correlation between personality and emotion.Aiming at the problem of missing personality text data and not making full use of prior knowledge,this paper proposes a personality detection model fused with knowledge graph.Firstly,the model filters the important concepts in the text through named entity recognition,and then combines the context information and the description information of the candidate entity to obtain the unique entity label associated with the candidate entity in the general domain knowledge graph to eliminate the entity ambiguity.Then,the entities and relations are mapped into a lowdimensional vector space through the knowledge graph embedding representation,and the corresponding knowledge representation is obtained.Finally,the semantic representation was obtained by combining knowledge representation and pre-trained model,and personality categories were detected by the machine learning classifiers.Experimental results show that the model introduces prior background information through the knowledge graph,enriches the semantic representation of the text,and achieves an average accuracy of 61.2%,which improves the accuracy of personality detection.Aiming at the problem that personality detection ignores psychological features and does not make full use of emotional information,this paper proposes a personality detection model based on emotion enhancement.Firstly,the model uses sentiment polarity analysis to filter the sentences with less obvious sentiment tendency in the input,and reduces the interference of irrelevant sentiment.Then,through fine-grained emotion recognition,the emotional intensity of the text in the four dimensions of joy,sadness,anger and fear is analyzed to obtain the corresponding emotional representation,which is combined with the semantic representation.Finally,the machine learning classifier is used to detect personality categories.In addition,the three features of knowledge,emotion and semantic are combined to complete the personality detection based on emotional knowledge enhancement.Experimental results show that the personality detection model based on emotion enhancement considers the correlation between emotion and personality,and uses emotional information to enrich the semantic representation of the text.The average accuracy reaches 62.6%.Compared with knowledge information,the gain of emotional information on personality detection is more obvious.Moreover,the performance of the personality detection model enhanced by emotional knowledge is particularly prominent,with an average accuracy of 63.8%,which is better than the current most advanced technology on each personality type.