Non-operative Personality Prediction Based on Knowledge Driven

Abstract

At present, most personality trait prediction studies mainly use the cooperative method, that is, using the scale to collect users' personality trait information. This method mainly has the disadvantages of strong subjectivity, limited quantity and quality, insufficient lasting stability and requiring users to cooperate. At the same time, the mainstream method uses the black box method of supervised learning, which belongs to the data-driven method and is not interpretable. Knowledge driven dictionary method is expected to solve these problems and realize non cooperative personality prediction. This paper proposes a method of constructing personality dictionary based on the combination of knowledge base and corpus. On the other hand, aiming at the unclear physical meaning of personality scoring algorithm in personality analysis using dictionary method, this paper proposes a personality scoring algorithm based on vocabulary weight and word frequency. The results show that the personality dictionary constructed by this method can ensure both timeliness and comprehensiveness in vocabulary. The experimental results show that the personality dictionary constructed by this method can ensure both timeliness and comprehensiveness in vocabulary. The average similarity between the predicted results of Weibo personality dictionary and the results of the scale is 61.98%, which is close to the results of BFM algorithm,which can effectively predict users' personality.