This paper describes intial expriments on word sense disambiguation (WSD) for the indonesian language. WSD is the task of determining the correct sense of a word according to the context it appers in. in these expriments, a number of polysemous words appering in the indonesian wordnet are randomly chosen, and google is used collect testing context paragraphs. some well-know vector model metrics are then applied, i.e. cosine similarity and singaular value decomposition (SVD) to solve the indonesian WSD problem. the results are compared against the human jugdements of three graduate students who were asked to select the most appropriate definition for each testing context paragraph. the expriment results showed that answers using cosine similarity achieved 62.5% similarity with human answers, whereas SVD achieved 67.5%. using the flesiss kappa statistic, cosine similarity-based WSD achieves an agreement of 0.311 with three human jugdes, whereas SVD is able to achieve.350.