Article: Chinese-English entity relation extraction method based on multi-layer semantic fusion and word information integration Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.18 pp.12 - 28 Abstract: Rapid growth in multilingual information has made cross-linguistic entity relationship recognition a critical NLP issue. However, existing methods often struggle with grammar differences, cross-linguistic semantic alignment, and structural feature utilisation, limiting robustness. Therefore, this study proposes a Chinese-English extraction method based on multi-layer semantic fusion and word information integration. This method constructs representations across word, phrase, sentence, and document levels. By integrating part-of-speech, dependencies, and location data via attention mechanisms, it deeply fuses semantic and structural features. Experimental results show that, in news corpora, the proposed method improves the accuracy of predicting entity relationships between Chinese and English by 2.99%, 6.63%, 5.10%, and 2.23%, respectively, compared to the other four methods. The method achieves a single-sample latency reduction of more than 12% in the forward inference stage of the model. Case studies further show a 0.94 prediction accuracy for 'investment project' categories with low confusion rates. Ultimately, this method balances accuracy, efficiency, and robustness, providing effective technical support for multilingual knowledge graphs and social computing. Inderscience Publishers - linking academia, business and industry through research

Title: Chinese-English entity relation extraction method based on multi-layer semantic fusion and word information integration

Authors: Lanlan Xiong

Addresses: General Education Teaching Department, Inner Mongolia Vocational College of Chemical Engineering, Hohhot, 010070, China

Abstract: Rapid growth in multilingual information has made cross-linguistic entity relationship recognition a critical NLP issue. However, existing methods often struggle with grammar differences, cross-linguistic semantic alignment, and structural feature utilisation, limiting robustness. Therefore, this study proposes a Chinese-English extraction method based on multi-layer semantic fusion and word information integration. This method constructs representations across word, phrase, sentence, and document levels. By integrating part-of-speech, dependencies, and location data via attention mechanisms, it deeply fuses semantic and structural features. Experimental results show that, in news corpora, the proposed method improves the accuracy of predicting entity relationships between Chinese and English by 2.99%, 6.63%, 5.10%, and 2.23%, respectively, compared to the other four methods. The method achieves a single-sample latency reduction of more than 12% in the forward inference stage of the model. Case studies further show a 0.94 prediction accuracy for 'investment project' categories with low confusion rates. Ultimately, this method balances accuracy, efficiency, and robustness, providing effective technical support for multilingual knowledge graphs and social computing.

Keywords: cross-linguistic entity relationship extraction; multi-layer semantic fusion; word information injection; attention mechanism; knowledge graph.

DOI: 10.1504/IJRIS.2026.155476

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.18, pp.12 - 28

Received: 26 Dec 2025
Accepted: 16 Mar 2026
Published online: 03 Aug 2026 *

Title: Chinese-English entity relation extraction method based on multi-layer semantic fusion and word information integration

Keep up-to-date