Can We Detect Bug Report Duplication with Unfinished Bug Reports?

Abstract

It is useful if a bug tracking system can detect bug report duplication with unfinished bug reports. To investigate the feasibility, we study relations between accuracy of duplicate bug report detection using features extracted from textual information in bug reports and the number of words in bug reports in this paper. The results show that increasing the number of words to be used in duplicate detection over a certain number does not affect the accuracy very much. The results also indicate that we had better use about 100 and 80 words in Eclipse and OpenOffice, respectively, in the detection because we may have many wrong candidates of duplication if we use words of more than the numbers. We thus think that detecting bug duplication in writing a new bug report has potential of giving duplicate bug report candidates.