| 杨阳蕊,朱亚萍,陈思思,刘雪梅,李慧敏.融合群体智能策略的AI链在大坝防汛抢险知识推理中的应用[J].水利学报,2023,54(9):1122-1132 |
| 融合群体智能策略的AI链在大坝防汛抢险知识推理中的应用 |
| Application of AI chain integrating crowd intelligence strategy in flood control knowledge inference |
| 投稿时间：2023-04-04 |
| DOI：10.13243/j.cnki.slxb.20230188 |
| 中文关键词: 防汛知识推理 防汛知识图谱 群体智能策略 AI链 |
| 英文关键词: flood control knowledge inference flood control knowledge graph crowd intelligence strategy AI chain |
| 基金项目:国家自然科学基金项目(72271091);河南省科学院科技开放合作项目(220901008);华北水利水电大学硕士创新能力提升工程项目(NCWUYC-2023091) |
|
| 摘要点击次数: 1631 |
| 全文下载次数: 1544 |
| 中文摘要: |
| 防汛抢险知识(实体、关系)是防汛抢险业务知识图谱的重要组成部分。防汛实体间关系错综复杂分布在无结构文本中，而可利用文本数量过少和文本质量偏低为该领域知识抽取工作带来了挑战。为此本文提出使用大型语言模型LLM(Large Language Model)进行大坝防汛抢险知识推理的思路。基于LLM设计防汛实体抽取器、防汛实体知识解析器以及防汛实体间关系决策器三个子模块，设计一系列有效任务提示，并将其链接形成人工智能AI (Artificial Intelligence) 链。通过AI链中提示与LLM实时交互逐步完成防汛知识推理任务。同时设计群体智能策略提高防汛实体间关系推理的可靠性。将本文提出的知识推理方法与现有方法进行对比，实验结果表明本文设计的AI链可有效提升LLM进行实体间关系推理的准确率，验证了AI链和群体智能策略的有效性。这一知识推理新范式可为提高水利防汛知识可访问性提供新的解决思路。 |
| 英文摘要: |
| Flood control knowledge (entities and relationships) is an essential component of the flood control and rescue business knowledge graph.The complex relationships among flood control entities are distributed in unstructured text，and the scarcity and low quality of available texts pose challenges to knowledge extraction in this field.In response，a novel approach is proposed to use large language models to complete flood control and rescue knowledge inference (FCRKI).Based on LLMs，three sub-modules are designed：flood control entity extractor，flood control entity knowledge parser，and flood control inter-entity relationship decider.A series of effective task prompts are designed and connected to form an AI chain.The flood control knowledge inference task is gradually completed through real-time interaction between the prompts and LLMs in the AI chain.Furthermore，a swarm intelligence strategy is designed to enhance the reliability of flood control inter-entity relationship inference.Comparing FCRKI with existing methods，experimental results show that FCRKI has a higher accuracy in inferring relationships between flood control entities，verifying the effectiveness of the AI chain and swarm intelligence strategy.This new knowledge extraction paradigm provides a novel solution for intelligent processing of hydraulic engineering texts. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |