Research Article
Research on Efficient Retrieval Method of Key Information of Discipline Inspection and Supervision Based on Digital Technology
@INPROCEEDINGS{10.4108/eai.3-11-2023.2342225, author={Tiancheng Chen and Yilin Cai}, title={Research on Efficient Retrieval Method of Key Information of Discipline Inspection and Supervision Based on Digital Technology}, proceedings={Proceedings of the 3rd International Conference on Innovation Design and Digital Technology, ICIDDT 2023, November 3--5, 2023, Zhenjiang, China}, publisher={EAI}, proceedings_a={ICIDDT}, year={2024}, month={1}, keywords={attribute characteristics;digital technology; efficient retrieval;information fusion;key information of discipline inspection and supervision; lightlda technology; likelihood model;}, doi={10.4108/eai.3-11-2023.2342225} }- Tiancheng Chen
Yilin Cai
Year: 2024
Research on Efficient Retrieval Method of Key Information of Discipline Inspection and Supervision Based on Digital Technology
ICIDDT
EAI
DOI: 10.4108/eai.3-11-2023.2342225
Abstract
Influenced by the characteristics of discipline inspection and supervision information itself, it is difficult to guarantee the correlation between the corresponding retrieval results and the retrieval objectives. Therefore, this paper puts forward the research on the efficient retrieval method of key information of discipline inspection and supervision based on digital technology. After comprehensively analyzing its specific attributes from four aspects: discipline inspection and supervision workflow information, discipline inspection and supervision object type information, discipline inspection and supervision content information and discipline inspection and supervision result information, the LightLDA technology in digital technology is introduced in the specific retrieval method design process, and the information in the retrieval likelihood model of discipline inspection and supervision information is fused with LDA by Wei and Croft in the LightLDA technology, and the information with the greatest likelihood to the retrieval information is taken as the retrieval output result. In the test results, the nDCG value is always stable above 40.0%,recall rates are consistently above 90%, which has obvious advantages compared with the control group.
