In the context of increasingly abundant digital network resources, existing N-gram language models face the problems of data scarcity and insufficient model generalization when processing text in the social media field. This paper aims to build and evaluate a language model system for digital network resources. Through data crawling technology, text data in related fields are collected from multiple digital platforms to ensure the diversity and representativeness of the data set; then text preprocessing methods, including word segmentation, stop word removal and word embedding technology, are used to convert the original data into a format suitable for model training; then a deep learning model based on the Transformer architecture is used for training, and hyperparameters are adjusted through cross-validation during the optimization process to improve model performance. Experimental results show that the model has an accuracy of 84.8% in specific domain texts, which verifies the effectiveness of the construction method. The proposed Transformer language model can not only improve the accuracy of text processing in specific domains, but also provide a reference framework and method for future language model research, which has important application value.