<![CDATA[Information retrieval in Indonesian folktales faces challenges in identifying semantically relevant stories, particularly due to language variation and synonymy. Keyword-based approaches such as TF-IDF often fail to capture semantic relationships, resulting in suboptimal search results. This study applies Latent Semantic Indexing (LSI) using Singular Value Decomposition (SVD) to enhance the relevance of story retrieval based on latent textual meaning. Evaluation was conducted through subjective assessments by 20 respondents, who rated search results generated from four types of queries: single-word, two-word, three-word, and full-sentence queries. The Friedman test indicated significant differences among query types for both LSI (χ²(3, N=20) = 29.27, p < 0.00001) and TF-IDF (χ²(3, N=20) = 35.45, p < 0.00001). LSI consistently yielded higher relevance scores, especially for more complex queries, demonstrating more effective semantic processing than TF-IDF. These findings suggest that semantic-based approaches are better suited for narrative texts such as folktales. Future research is recommended to explore query expansion and integration with deep learning-based semantic representations.]]>
Copyrights © 2025
