长江大保护工程经验知识抽取及策略智能生成
Knowledge extraction and strategic intelligence generation in the Yangtze River Protection Project
长江大保护工程采用EPC承包模式，属于涉及多利益相关者和复杂环境的大型工程。该项目面临复杂的协同关系和难度较大的协同管控挑战。因此，如何利用积累的经验知识实现高效、高质量辅助协同管控是各方面关注的焦点。为了有效利用文本知识，构建了BERT-BiLSTM文本分类模型和RoBERTa-BiLSTM-CRF实体识别模型。依托宜昌两网二期项目设计了知识抽取试验，进一步对长江大保护EPC项目文本资料进行知识抽取，经过实体对齐和结构化存储，最终形成了可复制、可推广的结构化协同策略库。此外，利用实体识别模型获得了问句关键实体信息，并通过匹配检索获得待定策略集。最后利用策略句融合方法，实现基于句子语义关系对待定策略集的语义融合，提出协同策略，为后续项目协同管控提供重要参考。
The Yangtze River Protection Project adopts the EPC contracting mode, which is a large-scale project involving multi-stakeholders and a complex environment. The project faces complex synergistic relationships and difficult synergistic control challenges. Therefore, how to utilize the accumulated empirical knowledge to realize efficient and high-quality assisted collaborative control is the focus of attention of all parties. In order to effectively utilize the text knowledge, BERT-BiLSTM text classification model and RoBERTa-BiLSTM-CRF entity recognition model were proposed. Knowledge extraction experiments were designed relying on the Yichang two-network phase II project, and further knowledge extraction was carried out on the textual information of the Yangtze River Great Protection EPC project, and after entity alignment and structured storage, a replicable and scalable structured collaborative strategy library was finally formed. In addition, the entity recognition model is utilized to obtain the key entity information of the interrogative sentence, and the pending strategy set is obtained through matching retrieval. Finally, the strategy sentence fusion method is utilized to realize the semantic fusion of the pending strategy set based on sentence semantic relations, and the synergistic strategy is proposed to provide an important reference for the synergistic control of subsequent projects.
/
| 〈 |
|
〉 |
京公网安备11010202008535号
PDF
(5061KB)
