Rahman and Roy proposed a novel technique BLIZZARD. This technique is primarily intended to address the challenge that bug reports may contain different types of structured information. Whether lacking rich structured information or having too much-structured information, can lead to poor search results. BLIZZARD determines whether there are too many program entities mentioned in the bug report and then reformulates the query to improve the accuracy of bug localization. BLIZZARD divides the bug reports into three categories according to different structural elements before query reformulation: stack traces, program elements, and natural language. Then correspond to three different graphs: Trace Graph, Text Graph and Source Term Graph. Experiments using 5,139 bug reports show that our technique can improve 59% of the noisy queries and 39% of the poor queries, which is promising.
Contributions of The Paper
• A novel query reformulation technique that filters out irrelevant information and adds relevant details to the bug report, and suggests improved queries for bug localization.
• A novel bug localization technique that locates bugs from the project source by employing a quality paradigm of bug reports, query reformulation, and information retrieval.
• Be a comprehensive evaluation of this new technique.
Comments
Provide more clear examples for three different types of bug reports in BLIZZARD. In the experiment, the author not only tested BLIZZARD in each bug group but also test in combined all types of bug reports, which can also prove that different techniques are only suitable for a certain type of bug report.
Publisher
Shiwen(Lareina) Yang
Link to The Paper
https://web.cs.dal.ca/~masud/papers/masud-ESECFSE2018.pdf
Name of The Authors
Mohammad Masudur Rahman, Chanchal K. Roy
Year of Publication
2018
Summary
Rahman and Roy proposed a novel technique BLIZZARD. This technique is primarily intended to address the challenge that bug reports may contain different types of structured information. Whether lacking rich structured information or having too much-structured information, can lead to poor search results. BLIZZARD determines whether there are too many program entities mentioned in the bug report and then reformulates the query to improve the accuracy of bug localization. BLIZZARD divides the bug reports into three categories according to different structural elements before query reformulation: stack traces, program elements, and natural language. Then correspond to three different graphs: Trace Graph, Text Graph and Source Term Graph. Experiments using 5,139 bug reports show that our technique can improve 59% of the noisy queries and 39% of the poor queries, which is promising.
Contributions of The Paper
• A novel query reformulation technique that filters out irrelevant information and adds relevant details to the bug report, and suggests improved queries for bug localization. • A novel bug localization technique that locates bugs from the project source by employing a quality paradigm of bug reports, query reformulation, and information retrieval. • Be a comprehensive evaluation of this new technique.
Comments
Provide more clear examples for three different types of bug reports in BLIZZARD. In the experiment, the author not only tested BLIZZARD in each bug group but also test in combined all types of bug reports, which can also prove that different techniques are only suitable for a certain type of bug report.