Document Retrieval Method Integrating Multi-Task Semantic Sorting Algorithm

Qi Liang ()
Additional contact information
Qi Liang: College of Computer Science and Cyber Security (Pilot Software College), Chengdu University of Technology, Chengdu 610059, P. R. China

Journal of Information & Knowledge Management (JIKM), 2025, vol. 24, issue 05, 1-19

Abstract: To improve the quality and efficiency of document retrieval, this study constructs a deep semantic recall model based on optimised bidirectional transformers in the recall stage of document retrieval, aiming to improve the matching degree of text semantics. In the sorting stage of document retrieval, a multi-task semantic sorting algorithm that integrates cross encoder and pairing method is proposed to achieve efficiency and accuracy in document retrieval. The experimental results show that the average accuracy of the recall model combined with contrastive learning is 0.773, while the average accuracy of the recall model further fused with deep semantics is 0.815. It indicates that the introduction of deep semantic information can significantly improve the semantic matching degree and model accuracy in the recall stage. The average accuracy of the proposed sorting model in five-fold cross validation was 0.801. The AUC values obtained by utilising the pairing method and the training method that combines the cross encoder and pairing method were 0.817 and 0.843. The query rates per second for the pipeline processes Det = 8, Rec = 2, batch = 1 and Det = 8, Rec = 2, batch = 4 were 46.6 and 47.2, respectively, with GPU utilisation rates of 94%. The application of research algorithms in document retrieval can improve user experience and retrieval efficiency with practical application meaning in document retrieval.

Keywords: Deep learning; OAG-BERT; Pair-wise semantic matching; semantic sorting algorithm; document retrieval (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.worldscientific.com/doi/abs/10.1142/S0219649225500480
Access to full text is restricted to subscribers

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:wsi:jikmxx:v:24:y:2025:i:05:n:s0219649225500480

Ordering information: This journal article can be ordered from

DOI: 10.1142/S0219649225500480

Access Statistics for this article

Journal of Information & Knowledge Management (JIKM) is currently edited by Professor Suliman Hawamdeh

More articles in Journal of Information & Knowledge Management (JIKM) from World Scientific Publishing Co. Pte. Ltd.
Bibliographic data for series maintained by Tai Tone Lim ().