A multi-layer dictionary learning algorithm that joints global constraints and Fisher discrimination (JGCFD-MDL) for image classification tasks was proposed. The algorithm reveals the manifold structure of the data by learning the global constraint dictionary and introduces the Fisher discriminative constraint dictionary to minimize the intra-class dispersion of samples and increase the inter-class dispersion. To further quantify the abstract features that characterize the data, a multi-layer dictionary learning framework is constructed to obtain high-level complex semantic structures and improve image classification performance. Finally, the algorithm is verified on the multi-label dataset of court costumes in the Ming Dynasty and Qing Dynasty, and better performance is obtained. Experiments show that compared with the local similarity algorithm, the average precision is improved by 3.34% . Compared with the single-layer dictionary learning algorithm, the one-error is improved by 1.00% , and the average precision is improved by 0.54% . Experiments also show that it has better performance on general datasets.

   How to protect cultural retics is of great significance to the transmission and dissemination of history and culture. Digital 3-dimensional (3D) modeling of cultural relics is an effective way to preserve them. The efficiency and complexity of cultural relic model reconstruction algorithms are significant challenges due to redundant data. To tackle the above issue, a 3D reconstruction algorithm, named COLMAP + LSH, was proposed for movable cultural relics based on salient region optimization. COLMAP + LSH algorithm introduces saliency region detection and locality-sensetive Hashing (LSH) to achieve efficient, accurate, and robust digital 3D modeling of cultural relics. Specifically, 400 cultural model data were collected through offline and online collection. COLMAP + LSH algorithm detects the salient region interactively and reduces the number of images in the salient region by feature diffusion. Additionally, COLMAP + LSH algorithm utilizes LSH to calculate the image selection scores and employs the image selection scores to reduce the redundant image. The experiments on the self-constructed cultural relics dataset show that COLMAP + LSH algorithm can efficiently achieve image feature diffusion and ensure the quality of artifact reconstruction while selecting most of the redundant image data.

   Few-shot named entity recognition (NER) aims to identify named entities in new domains using a limited amount of annotated data. Previous methods divided this task into entity span detection and entity classification, achieving good results. However these methods are limited by the imbalance between the entity and non-entity categories due to the use of sequence labeling for entity span detection. To this end, a point-proto network ( PPN) combining pointer and prototypical networks was proposed. Specifically, the pointer network generates the position of entities in sentences in the entity span detection stage. The prototypical network builds semantic prototypes of entity types and classifies entities based on their distance from these prototypes in the entity classification stage. Moreover, the low-rank adaptation ( LoRA) fine-tuning method, which involves freezing the pre-trained weights and injecting a trainable decomposition matrix, reduces the parameters that need to be trained and saved. Extensive experiments on the few-shot NER Dataset (Few-NERD) and Cross-Dataset demonstrate the superiority of PPN in this domain.

   As a subtask of open domain event extraction ( ODEE), new event type induction aims to discover a set of unseen event types from a given corpus. Existing methods mostly adopt semi-supervised or unsupervised learning to achieve the goal, which uses complex and different objective functions for labeled and unlabeled data respectively. In order to unify and simplify objective functions, a reliable pseudo-labeling prediction (RPP) framework for new event type induction was proposed. The framework introduces a double label reassignment ( DLR) strategy for unlabeled data based on swap-prediction. DLR strategy can alleviate the model degeneration caused by swap-predication and further combine the real distribution over unseen event types to produce more reliable pseudo labels for unlabeled data. The generated reliable pseudo labels help the overall model be optimized by a unified and simple objective. Experiments show that RPP framework outperforms the state-of-the-art on the benchmark.