信息工程与人工智能学院知识库

随着人口激增和经济的快速发展，气候变化和资源枯竭等全球性危机日益加 深，人们对社会、经济、生态三个领域的可持续发展越来越重视，特别是在自然 生态环境领域的可持续发展研究一直是学术研究的热门议题。同时祁连山作为我 国西部重要的生态安全屏障和重要水源产地，祁连山地区的可持续发展对于我国 西部乃至全国的生态文明建设至关重要。当前由于气候环境的变化和人类活动的 影响，祁连山面临着日益严重的环境问题和可持续发展挑战，对祁连山可持续发 展现状进行成体系的归纳整理和研究成为重要而紧迫的任务。当下自然语言处理、 深度学习技术的飞速发展，为研究处理祁连山可持续发展信息提供了新的方向， 知识图谱的出现也为相关生态环境学术研究和可持续发展领域提供了直观有效 的工具。因此，本文利用知识图谱思想，构建祁连山可持续发展知识图谱，为祁 连山可持续发展的智能研究、知识问答、知识推理等方面提供重要支撑。本文主 要内容如下： （1）建立祁连山可持续发展信息数据集。将中国知网数据库中公开发表的 中文期刊论文作为主要数据源，检索祁连山可持续发展相关信息文献并进行处理， 建立原始数据集用于祁连山可持续发展文献研究热点分析，确定好实体类型和关 系类型后对文本数据进行标注，最后把文本转换成特定格式完成标注数据集的创 建。标注数据集共包含6种实体类别和6种关系类别，为后续构建知识图谱提供 数据支撑。 （2）提出了融合注意力机制的 ALBERT-BiLSTM-Attention-CRF 轻量化模 型。模型在特征提取层BiLSTM的后面引入注意力层，解决了BiLSTM模型存 在的问题，同时轻量化的ALBERT层使得模型在本文的命名实体语料规模相对 较小的情况下，以其参数少的优势，能够取得更好的性能。经过与其他命名实体 识别模型实验对比，本文模型的准确率、F1 值都有较高的提升，证明了模型的 有效性与可行性。 （3）构建祁连山可持续发展知识图谱。利用确定好的实体和关系类型，通 过 ALBERT-BiLSTM-Attention-CRF 模型进行实体识别，根据抽取的实体特征， 基于规则模版完成实体关系抽取，构建<实体，关系，实体>三元组，最后采用 Neo4j 图数据库对知识图谱进行存储与可视化展示。

With the rapid population growth and rapid economic development, global crises such as climate change and resource depletion are deepening, and people are paying more and more attention to sustainable development in the social, economic, and ecological fields. In particular, the research on sustainable development in the natural ecological environment has always been a hot topic in academic research. At the same time, as an important ecological security barrier and important water source in western China, the sustainable development of the Qilian Mountains is crucial for the construction of ecological civilization in western China and even the whole country. Currently, due to changes in climate environment and the impact of human activities, the Qilian Mountains are facing increasingly serious environmental problems and sustainable development challenges. It has become an important and urgent task to systematically summarize and study the current situation of sustainable development in the Qilian Mountains. The rapid development of natural language processing and deep learning technology provides a new direction for researching and processing sustainable development information in the Qilian Mountains, and the emergence of knowledge graphs also provides intuitive and effective tools for related ecological environment academic research and sustainable development fields. Therefore, this article uses the idea of knowledge graphs to construct a knowledge graph of sustainable development in the Qilian Mountains, providing important support for intelligent research, knowledge question answering, knowledge reasoning, etc. in sustainable development of the Qilian Mountains. The main contents of this article are as follows: (1) Establish a sustainable development information dataset for the Qilian Mountains. Using the Chinese journal articles published in the CNKI database as the main data source, retrieve and process the relevant information literature on sustainable development in the Qilian Mountains, establish an original data set for the analysis of research hotspots in sustainable development literature in the Qilian Mountains, and label the text data after determining the entity types and relationship types. Finally, convert the text into a specific format to complete the creation of labeled data sets. The labeled data set contains six types of entities and six types of relationships, providing data support for the subsequent construction of a knowledge graph. (2) A lightweight model of ALBERT-BiLSTM-Attention-CRF with attention mechanism is proposed. The model introduces the attention layer behind the feature extraction layer BiLSTM, which solves the problems of BiLSTM model. At the same time, the lightweight ALBERT layer enables the model to achieve better performance with less parameters in the case of relatively small scale of the named entity corpus in this paper. Compared with other named entity recognition models, the accuracy of the proposed model The F1 value has a higher increase, which proves the effectiveness and feasibility of the model. (3) Construct a knowledge graph for sustainable development in the Qilian Mountains. Using the determined entity types and relationship types, perform entity recognition through the ALBERT-BiLSTM-Attention-CRF model. Based on the extracted entity features, complete entity relationship extraction based on rule templates, construct a triplet, and finally use the Neo4j graph database to store and visualize the knowledge graph.