The Intersection of Linguistics and Information Technology in the Global Communication of Martial Arts Culture

Abstract

In the course of global dissemination, Wushu culture encounters numerous linguistic challenges that significantly impede the translation quality and dissemination efficacy of its content. These challenges include terminology barriers, cultural heterogeneity, linguistic style conventions, and audience applicability. This paper initially explores the representation of natural language within computational systems and examines the foundational frameworks of language models. Subsequently, we algorithmically model the basic architecture of neural machine translation and text feature extraction. To address the gradient error issue in backpropagation, we employ the Long Short-Term Memory (LSTM) algorithm. Furthermore, we construct a neural machine translation model that incorporates an enhanced attention mechanism within the LSTM framework. To refine this model, we implement data augmentation techniques, aiming to optimize performance and improve translation accuracy in the context of Wushu's multifaceted linguistic landscape. On this basis, the translation effect of martial arts content is tested and simultaneously used to retranslate 40 pieces of the hottest overseas martial arts culture content. The quality of the translated text before and after retranslation is analyzed to analyze the enhancement effect of the model on the impact of the global dissemination of martial arts culture. It is found that the model in this paper has the best performance in both directions of English-to-Chinese and Chinese-to-English translation, with precision and recall scores of (0.908,0.876) and (0.915,0.922), and F1 values of 0.869 and 0.844, which make the performance stable. In the retranslation and translation quality test of popular martial arts video content, compared with the total penalty score of 1013.7 points before retranslation, the penalty score of the sample after retranslation totaled 246.2 points, a significant drop of 767.5 points, and this paper's model has a substantial effect on the retranslation of martial arts content. This study presents research ideas and practical methods from the perspective of information technology to enhance the global communication capabilities of martial arts culture and establish the international influence of Chinese culture.

Keywords

• machine learning
• rnn
• lstm
• corpus
• martial arts culture
• 68-02