Article: Semantic analysis-integrated LSTM neural network: a novel spelling correction approach for academic English translation Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.68 pp.27 - 45 Abstract: To address the issue of existing spelling correction methods neglecting semantic relevance in academic English translation quality analysis, a semantic spelling correction algorithm is proposed. The pipeline first performs pre-analytic validation to reduce spelling and semantic noise, then conducts coherence, grammaticality, and terminology assessments using learned features. A curated corpus comprising 20,000 academic text fragments is utilised for training and evaluation, and benchmark baselines are included to ensure methodological comparability. Quantitative results demonstrate that the designed algorithm achieves a spelling error recognition rate of 93.45% and a processing speed of 240.78 words/second, significantly improving the accuracy and efficiency of spelling correction, while maintaining semantic integrity (cosine similarity 0.87), which is significant for improving the quality of academic English translation. The work reframes correction as methodological infrastructure within quality analysis, integrating a semantic-aware module that safeguards metric fidelity before analytic scoring. Inderscience Publishers - linking academia, business and industry through research

Title: Semantic analysis-integrated LSTM neural network: a novel spelling correction approach for academic English translation

Authors: Hui Gao; Jing Wu

Addresses: Beijing University of Financial Technology, Beijing, 101118, China ' Beijing University of Financial Technology, Beijing, 101118, China

Abstract: To address the issue of existing spelling correction methods neglecting semantic relevance in academic English translation quality analysis, a semantic spelling correction algorithm is proposed. The pipeline first performs pre-analytic validation to reduce spelling and semantic noise, then conducts coherence, grammaticality, and terminology assessments using learned features. A curated corpus comprising 20,000 academic text fragments is utilised for training and evaluation, and benchmark baselines are included to ensure methodological comparability. Quantitative results demonstrate that the designed algorithm achieves a spelling error recognition rate of 93.45% and a processing speed of 240.78 words/second, significantly improving the accuracy and efficiency of spelling correction, while maintaining semantic integrity (cosine similarity 0.87), which is significant for improving the quality of academic English translation. The work reframes correction as methodological infrastructure within quality analysis, integrating a semantic-aware module that safeguards metric fidelity before analytic scoring.

Keywords: academic English translation; spelling correction algorithm; semantic analysis; academic quality assessment; deep learning model.

DOI: 10.1504/IJICT.2026.154349

International Journal of Information and Communication Technology, 2026 Vol.27 No.68, pp.27 - 45

Received: 28 Nov 2025
Accepted: 12 Jan 2026
Published online: 23 Jun 2026 *

Title: Semantic analysis-integrated LSTM neural network: a novel spelling correction approach for academic English translation

Keep up-to-date