A graph convolutional network based on object relationship method under linguistic environment applied to film evaluation

Film evaluation is of considerable significance to the development of the film and TV indus-try. A film evaluation is usually a qualitative evaluation. Existing research focuses on trans-forming qualitative evaluation into numerical values and analyzing numerical values. However, existing methods have a semantic loss in the quantization process, and perfor-mance degradation occurs in the mass data. In this paper, a graph convolutional network based on object relationships (OR-GCN) is proposed under linguistic environment and applied to film classification and ranking. First, the dominant matrix is obtained according to the evaluation between objects, and the object relationship is constructed by using the dominant matrix. Second, the graph convolutional network is used to extract the object relationships, deeply learned the relationship between objects, and classified and sorted the objects. Finally, on film review data of Douban (douban.com), the films are classified and sorted by the OR-GCN model, and the effectiveness and the non-randomness of this method are verified by the accurate analysis and ROC. At the same time, our method is applied to the public dataset to illustrate the performance and universality. In this paper, the proposed OR-GCN model can avoid linguistic quantization and only consider the rela-tionship between objects, and provide a new perspective for solving language term set problems. (c) 2022 Elsevier Inc. All rights reserved.