A Comparative Study of the Performance of IR Models on Duplicate Bug Detection
Bibliographic record
Abstract
Open source projects incorporate bug triagers to help with the task of bug report assignment to developers. One of the tasks of a triager is to identify whether an incoming bug report is a duplicate of a pre-existing report. In order to detect duplicate bug reports, a triager either relies on his memory and experience or on the search capabilities of the bug repository. Both these approaches can be time consuming for the triager and may also lead to the misidentification of duplicates. In the literature, several approaches to automate duplicate bug report detection have been proposed. However, there has not been an exhaustive comparison of the performance of different IR models, especially with topic-based ones such as LSI and LDA. In this paper, we compare the performance of the traditional vector space model (using different weighting schemes) with that of topic based models, leveraging heuristics that incorporate exception stack frames, surface features, summary and long description from the free-form text in the bug report. We perform experiments on subsets of bug reports from Eclipse and Firefox and achieve a recall rate of 60% and 58% respectively. We find that word-based models, in particular a Log-Entropy based weighting scheme, outperform topic based ones such as LSI, LDA and Random Projections. Our findings also suggests that for the problem of duplicate bug detection, it is important to consider a project's domain and characteristics to devise a set of heuristics to achieve optimal results.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.031 | 0.061 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.003 | 0.002 |
| Bibliometrics | 0.005 | 0.004 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.006 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".