logseq/pages/COLA-D-23-00096.md

type:: REVIEWS tags:: year:: 2023 venue:: COLA full-title:: AI-based Clustering of Similar Issues in GitHub’s Repositories date-start:: 20-09-2023 - 15:24 date-submitted:: external-links:: status:: DONE deadline-submission:: 31-10-2023 file::

- [[Highlights]]
	- #+BEGIN_IMPORTANT
	  ((654cf595-c370-437b-9b2f-dabaf84d8c0e))
	  ((654cf5c4-718b-4d37-aaea-b4658b5ed7ad))
	  ((654cf5d6-940f-4078-a6fe-65af2bca9afb))
	  #+END_IMPORTANT
	- #+BEGIN_WARNING
	  ((654d0b5e-8aa3-4eb4-ac26-83e9fc224d13))
	  *To be checked*
	  #+END_WARNING
	- ((6554f0d6-ae81-4f8c-bfab-a7041da86d96))
		- The clustering is done by considering title and body of each issue. ==Why not considering also labels?== Have you evaluated that labels do not have any impact on the results?
			- This is the answer:
				- ((654d0f20-3e2b-4b6d-bddd-015d4a3ad297))
		- ((6554f1eb-38bf-447e-85dc-b41d6594246e))
			- What's the similarity value that is used as threshold?
				- That's the answer:
					- ((6554f27b-e237-409f-ae81-1966cf5cac44))
	- ((654d1026-9678-4b52-b875-57d0c7e594a0))
		- ==This is a huge bias. The difficult part here is to identify issues that are similar but not the same.==
	- ((65537124-859a-4460-991f-5652b10380e5))
		- Diverse with respect to what?
	- ((65537619-ffb9-4fd6-aea0-a46bd19bb3b6))
		- -> Credibility?
	-
	-
	-
	- ### Minors
		- ((654d098f-d572-4f9d-af6e-d8c141179f66))
			- -> GutHub
		- ((654d0a15-8279-4622-85d1-a8d2abb01f6d))
			- Improve the quality of the Figure
		- ((654d0a7d-0b3b-4802-8d5b-dcc99ac20fe1))
			- Commas should be at the apex
		- ((654d0aa6-3ed1-4ec6-85a0-548508d471ae))
			- The figure is not in the right ratio
		- ((654d0f3f-322b-4db1-9b86-509a6d740969))
	-
- [[Comments]]
	- Summary: The paper introduces a machine learning methodology aimed at efficiently grouping similar issues on GitHub, streamlining the assignment process and facilitating faster issue resolution. The proposed approach calculates textual similarity among issues within a specific repository, identifying cohesive clusters of related issues with promising outcomes. Future investigations will explore additional issue features and semantic similarity measures.
	- Comments: The paper addresses an intriguing and pertinent topic, effectively presenting the problem and detailing the proposed solution. However, several concerns arise regarding the evaluation process:
		- The evaluation centers on a dataset comprising duplicate clusters of issues. While the assertion that "100% duplicate issues are similar issues, too" is acknowledged, this poses a significant threat to the validity of the evaluation, as it is conducted within a narrowly defined context. This focus on duplicates may not sufficiently capture the broader real-world scenarios where issue similarity is more nuanced.
		- The paper needs more clarity regarding the treatment of new issues introduced during the evaluation. It is essential to reapply the Agglomerative Hierarchical Clustering methodology to gauge the impact of these additional issues.
		- Given that the dataset primarily consists of duplicated issues, a qualitative discussion is warranted to explain why precision and recall values are not consistently high across projects, with a minimum recall value potentially as low as 0.25.
		- While Research Question 4 is well-formulated, the paper falls short in providing a satisfactory answer. To effectively assess the practical utility of the approach and its impact on reducing human effort in issue identification, a combination of quantitative and qualitative user studies is essential for a comprehensive response to the research question.
	-
	-
- [[REVIEWS/Notes]]