摘要：Textual documents have been largely available digitally or electronically with increasing number since the invention of computer. The need of information makes a user to do a searching activity on document collections he or she has. Document searching method available on computers usually just look into the name of document (file) searched by user without noticing the content inside it, therefore it can not satisfy his or her information need. Latent Semantic Indexing (LSI) method can be a solution to retrieve relevant information from document collections inside a computer. LSI method has an ability to correlate semantically related terms that are latent in a collection of text using Singular Value Decomposition (SVD) technique. An information retrieval system using LSI method has been developed and tested on 40 Indonesian documents with 478 words per document in average. On testing process which has been done, with threshold of 0.6 the system could give search result with average precision score of 55.48% and average recall score of 96.67% in testing process.
关键词：information retrieval, latent semantic indexing, singular value decomposition, precision, recall.
