A Study on Semantic Default of Culture-Loaded Words in Intangible Cultural Heritage Folk Songs Empowered by AI

Taking culture-loaded words in intangible cultural heritage (ICH) folk songs as the research object, this paper explores the semantic default phenomenon in artificial intelligence (AI) translation. Based on Nida’s Functional Equivalence Theory, this study focuses on ICH folk songs in northern and southern China, and analyzes translated examples of material culture-loaded words, social relation culture-loaded words, and ideological and philosophical culture-loaded words with representative works such as The Sun Rises Joyfully and Going to the West Gate. Taking DeepSeek as a case, the analysis finds that AI encounters difficulties in processing various types of culture-loaded words: due to the lack of cultural background and lexical gaps, the translated versions fail to effectively convey cultural images, emotional connotations, and aesthetic experiences. To improve the quality and efficiency of AI translation, this paper proposes to enhance AI’s ability to recognize and extract cultural elements by constructing a high-quality cultural corpus.