Airborne products with diverse range and extensive industrial chain have a complex testing system， requiring comprehensive metrological verification work. However， airborne product data resources primarily exist in unstructured， fragmented， and multimodal forms， making it difficult to conduct overall analysis of various testing elements or trace the standardization of testing and product quality under a unified framework， thereby posing challenges to metrological work. To address this issue， the construction of knowledge graph for Metrological Traceability of Airborne Products （MT-AP） was explored by combining generative Large Language Model （LLM）. Firstly， the resource types and metrological traceability links were sorted out， and a Knowledge Graph （KG） ontological model was constructed. Secondly， LLM-based work modules were designed and integrated into workflow chains. Finally， a method for constructing the MT-AP knowledge graph based on the workflow chains and prompt templates was proposed. Experiments were conducted using airborne product instance data and workflow chains. Experimental results show that the proposed method has the knowledge comprehension and naming capability scored above 0.91 basically， the text segmentation and knowledge decoupling capability scored above 0.83 basically， and the complex parameter extraction and structured capability scored above 0.85 basically. It can be seen that the proposed method exhibits satisfactory performance in key tasks of MT-AP knowledge graph construction， providing technical support for metrology engineering of airborne products.

With the development of Pretrained Language Model （PLM）， Retrieval-Augmented Generation （RAG） is widely concerned as an emerging task. A comprehensive and objective evaluation of RAG is considered essential to reveal the limitations of the existing methods and to indicate future research directions. However， a lack of systematic evaluation benchmarks for RAG is observed， especially in the context of long documents. To address this issue， an automatic question-answering construction strategy based on focused fragments was proposed， aiming to build large-scale QA datasets efficiently and accurately. Based on this strategy， the first bilingual RAG evaluation benchmark dataset for long documents， named LoRAG， was constructed， covering English-Chinese bilingual documents from multiple domains such as law， finance， and literature， with an average document length of 57 000 tokens in English and 76 000 tokens in Chinese. Systematic experiments on the two key stages of RAG — retrieval and generation were conducted using the LoRAG dataset. In the retrieval stage， multiple mainstream embedding models， including text-embedding-ada-002， the bge-large series， bge-m3， and Multilingual-E5-large-instruct， were evaluated， and the reranking model bge-reranker-v2-m3 was introduced for performance optimization and comparison. In the generation stage， representative Large Language Models （LLM）， including Vicuna-13B， ChatGLM2-6B， Llama2-7B， and Claude2， were tested comprehensively. Experimental results show that the constructed dataset LoRAG reveals the positioning challenges faced by current embedding methods in long-document retrieval， as well as the limitations of LLM in balancing relevance and conciseness during the generation process， providing clear research directions for future method improvements.

In recent years， Large Language Models （LLMs） have demonstrated excellent language understanding and dialogue capabilities in fields such as natural language processing and computer vision. However， they can produce inference results that are inconsistent with the correct answers in professional fields. This situation brings significant challenges to the application of LLMs in precise and accurate decision-making tasks. To solve this problem， a rule-guided Post Prompt of Large Language Model （PP-LLM） generation method was proposed. In this method， by generating post prompts， the original problem was transformed into two sub-problems that are easier to solve， thereby achieving the purposes of introducing expert knowledge and reducing the difficulty of task learning. Specifically， the knowledge-guided specific rules were used to transform the output part of the supervised dataset into a combination of post prompts and the output portion. PP-LLM method does not change the training and inference processes of the model， and does not add computational cost. Experimental results show that PP-LLM method significantly improves the accuracy of inference results and narrows the gap between model predictions and actual answers. Compared with the results without using the proposed method， the F1 value and Recall-Oriented Understudy for Gisting Evaluation （ROUGE） of the PP-LLM method have significantly improved. It can be seen that the above work improves the reliability of LLMs in professional applications and provides new ideas for LLM generation technology.

Addressing the insufficient exploration of item-transition information within the current session and the limited utilization of other session details in session-based recommendation nowadays， a multi-layer information interactive fusion algorithm based on graph neural network was proposed for session-based recommendation. Based on the current session， firstly， the information of neighborhood nodes was aggregated by designing different weights for the connection relationships between nodes， and the explicit information of item-transition in the current session was mined. Secondly， the neighborhood node information was aggregated by stacked residual graph attention network， and the implicit item-transition information in the current session was mined. Finally， the sequence-dependent information in the time stamp-based session was mined through a single gated graph neural network. Based on other sessions， the entire set of sessions was linked through the first-order neighbors of nodes， and the global information encoding was learnt， and then， the embedding representations of four levels were integrated to obtain more comprehensive item-transition information. At the same time， soft attention mechanism and reverse position embedding information were used to fuse the obtained item-transition information more effectively. Experimental results show that the precision P@20 and mean reciprocal rank MRR@20 of the proposed algorithm are increased by 0.79% and 0.84% respectively compared with the suboptimal model GCE-GNN （Global Context Enhanced Graph Neural Network） on Diginetica dataset， the P@20 and MRR@20 of the proposed algorithm are increased by 8.23% and 7.86% respectively compared with the suboptimal model HyperS²Rec on Tmall dataset， and the P@20 and MRR@20 of the proposed algorithm are increased by 1.33% and 7.16% respectively compared with the suboptimal model HyperS²Rec on Nowplaying dataset.

At present， in skeleton-based action recognition task， there still are some shortcomings， such as unreasonable data preprocessing， too many model parameters and low recognition accuracy. In order to solve the above problems， a skeleton-based action recognition method based on feature interaction and adaptive fusion， namely AFFGCN（Adaptively Feature Fusion Graph Convolutional Neural Network）， was proposed. Firstly， an adaptive pooling method for data preprocessing to solve the problems of uneven data frame distribution and poor data frame representation was proposed. Secondly， a multi-information feature interaction method was introduced to mine deeper features， so as to improve performance of the model. Finally， an Adaptive Feature Fusion （AFF） module was proposed to fuse graph convolutional features， thereby further improving the model performance. Experimental results show that the proposed method increases 1.2 percentage points compared with baseline method Lightweight Multi-Information Graph Convolutional Neural Network （LMI-GCN） on NTU-RGB+D 60 dataset in both Cross-Subject （CS） and Cross-View （CV） evaluation settings. At the same time， the CS and Cross-Setup （SS） evaluation settings of the proposed method on NTU-RGB+D 120 dataset are increased by 1.5 and 1.4 percentage points respectively compared with those of baseline method LMI-GCN. And the experimental results on single-stream and multi-stream networks show that compared with current mainstream skeleton-based action recognition methods such as Semantics-Guided Neural network （SGN）， the proposed method has less parameters and higher accuracy of the model， showing obvious advantages of the model， and that the model is more suitable for mobile device deployment.