[Objective] Aiming at the problems of wide differences and lack of standards in the value evaluation of calligraphy works, this paper used big data and artificial intelligence methods to efficiently and accurately identify the value elements of calligraphy works and provide technical support for various calligraphy works trading activities.

[Methods] Combining hyperplane algorithm and Bert model to preprocess calligraphy documents by stop words screening and semantic expansion to form an optimized corpus with high recognition; Constructing complex semantic network of calligraphy literature and introducing Louvain algorithm to determine the optimal number of topics by maximizing the modularity of community network. Therefore, this paper put forward a new method based on "Hyperplane -Bert-Louvain-LDA" (HBL-LDA) to efficiently and accurately construct the evaluation index system of calligraphy value.

[Results] Experiments on calligraphy value literature showed that compared with the traditional LDA model, the precision and F value of the topic recognition of the HBL _ LDA model were increased by 45 % and 29.46 % respectively. What’s more, the average topic quality rate was less 0.96085, and more high-quality topics were identified. Finally, based on the representative calligraphy works, a variety of regression models were used to verify the evaluation index system, and the regression decision tree had the highest accuracy of 84 %.

[Limitations] The new model only constructed an evaluation index system for calligraphy works. In the future, it would incorporate multi-source data of other works of art into the construction of cultural product value indicators. Moreover, because the Bert model don’t consider the topic semantic information, the expansion of similar feature words had certain limitations.

[Conclusions] In this paper, a new model of calligraphy value evaluation index system based on "hyperplane -Bert-Louvain combination optimization LDA model" was proposed, which provided a new direction for the construction of index system in other fields. The index system constructed in this paper was easy to operate and adaptable, and could quickly generate a new index system as demand changes.