The theory of “Bie-modernist Cultural Computing” was proposed by Professor Wang Jianjiang, which organically combines Bie-modernism with cultural computing and aesthetics with computers, and implements Bie-modernism in human-computer interaction and cultural computing. It is a form of cultural computing guided by the concept of Bie-modernism. “Bie-modernism” is a theory about social forms and historical development stages, and has sparked intense discussions in both domestic and international academic circles. The “Bie-modern” in Bie-modernism distinguishes between true and false phenomena, and Bie-modernism is a reflection and criticism of Bie-modern phenomena. Bie-modernist Cultural Computing is the integration of Bie-modernism and cultural computing into the field of computer artificial intelligence, converting the concept of distinguishing true and false in Bie-modernism into Bie-modernist Cultural Computing, achieving a leap in Bie-modernism theory and having a broader prospect in the virtual space. This article applies Bie-modernist Cultural Computing and computer metaphor computing to the analysis of the imagery in Li Shangyin’s poetry. Metaphor is an important thinking and means adopted by Li Shangyin in his poetry creation. By using this thinking and means, the poet concretizes a large number of abstract and difficult-to-understand concepts, allowing the appreciators to delve into the inner part of the poetry from the external appearance and understand the deeper meaning of the poetry and the viewpoints and emotions the poet intends to express. Metaphor is both an ancient and a novel topic. Its research originated in the ancient Greek-Roman period and was confined to the realm of rhetoric. In the modern period, it has flourished. In the second half of the 20th century, the study of metaphor in disciplines such as cognitive science, psychology, linguistics, philosophy, literature, and aesthetics has reached an unprecedented high, showing a trend of continuous integration among these disciplines. Li Shangyin’s poetry often features complex imagery, ambiguous moods, and rich colors and flavors. The expression and interpretation of his works have great tension, posing significant challenges to the interpretation of his poems. Therefore, the computer metaphor recognition system also faces certain difficulties in interpreting his poems. Many of Li Shangyin’s poems generate new metaphorical meanings through the interaction between the subject and the object of the metaphor, based on the poet’s immediate experiences and feelings. Besides grasping the characteristics of similarity metaphorical imagery from the process of metaphor occurrence, we can also explore their features from the perspective of function. In Li Shangyin’s poetry, the interaction between the subject and the object of the metaphor generates similarity, which is based on the inherent affinity of the same kind within the poem. The foundation of the metaphorical ends in Li Shangyin’s poetry imagery is their similarity. However, often the subject and the object have no relation at all, making it difficult to find the metaphorical base. This indicates that the metaphorical foundation established between the subject and the object is a new and creative similarity, that is, dissimilarity. The application of Bie-modernist Cultural Computing to the field of computer metaphor computing is of pioneering significance. Metaphor computing research has become a hot topic in the field of natural language processing. Neural networks have demonstrated superior performance in many natural language processing tasks, so using neural networks for metaphor computing has advantages. The feedforward neural network is the simplest form of neural network. The basic unit of the feedforward neural network is the neuron. The Convolutional Neural Network (CNN) was first used in computer vision and mainly consists of three parts: the word vector layer, the convolution layer, and the pooling layer. Currently, in the field of natural language processing research, metaphor computing is divided into three sub-tasks: metaphor recognition, metaphor understanding, and metaphor generation. Metaphor recognition refers to detecting sentences that form metaphorical expressions in the text and marking them; metaphor understanding refers to providing an interpretable description of metaphorical sentences; metaphor generation refers to the ability of the computer to automatically generate metaphorical text sentences through certain algorithms. The essence of metaphor recognition is to find the contradictory relationship between semantics. Due to the complexity and diversity of metaphors, researchers from related disciplines have designed different recognition algorithms for different metaphor categories. Currently, the more common algorithms in metaphor recognition can be divided into four main categories: recognition algorithms based on textual cues, semantic knowledge, machine learning, and neural networks. Metaphor recognition is the fundamental research for metaphor understanding and generation. Only when the accuracy of the metaphor recognition model reaches a certain threshold and the model has strong stability can the two tasks of metaphor understanding and generation be developed and studied on a large scale. It is obvious that in order for computers to provide explainable descriptions of metaphorical expressions and to automatically generate metaphorical text sentences, the first step is to achieve metaphor recognition research. Relying on manually constructed metaphor knowledge bases, it is difficult to build metaphor knowledge bases for massive data. Neural network models have strong classification and representation capabilities and have been frequently used to handle metaphor recognition in recent years. However, the recognition accuracy of metaphor recognition methods based on machine learning still has a lot of room for improvement. Therefore, the combination of neural network models and postmodernist cultural computing can be more precise and in-depth.