This study examines Latent Semantic indexing (LSI) using Singular Value Decomposition (SVD) in the knowledge retrieval process, namely indexing Indonesian text. There are three stages in this process: (1) text processing, which consists of tokenisation, filtering, and stemming process, (2) developing LSI using SVD and (3) evaluating and measuring performance. The result showed Mean Average Precision around 77.90% on scenario matrix dimension 120 and average precision for first retrieval around 83.33% on scenario matrix dimension 90.