Performance evaluation of VSM and LSI models to determine bug reports similarity

Bug reports of open source software systems are increasing exponentially. One reason for growing bug reports is that bug reporters do not browse the bug repository before submitting a bug report. There may be some similar bugs already reported: one, which are exactly similar or duplicate and other, which are semantically similar means they may belong to the same software component or files. The information contained in the previously reported similar bugs can be helpful in fixing and resolving the newly reported bugs. In this paper, we applied and compared performance of two information retrieval (IR) models: Vector Space Model (VSM) and Latent Semantic Indexing (LSI), in extracting existing similar bug reports. The performance of these two models have been evaluated based on the Top Ten results retrieved by them for relevant bug reports. Experiments have been conducted on 106 bug reports of three components from Google chrome, browser. Result shows that LSI performs better in most cases in comparison to VSM.