Using Computer-Based Word Processing Methods to Analyze the Quality of Bug Reports

Software companies incur large costs for fixing bugs in their product. Systems for collecting errors are supplemented by bugs that are written by different people, with different levels of knowledge, skills and time availability. Often they cannot figure out whether this problem was already fixed or not, as a result, they start duplicate error reports. Error fixing systems are often crowded with the same error reports. Despite this, each report should be thoroughly examined for possible errors. And the person responsible for the processing of recently opened reports should group them, so that the developers do not spend too much time on repeated work. Finding duplicates is becoming quite an expensive process. In this work, we consider text classification methods to reduce the time spent on this process. The thesis discusses approaches for recommending similar error reports. Approaches combine the traditional technique of finding information and the technique of embedding words. They accept headers and descriptions of errors as well as product version and error priority. To evaluate the approaches, a dataset from an open source Eclipse project is used.