Who is who in the mailing list? Comparing six disambiguation heuristics to identify multiple addresses of a participant
Many software projects adopt mailing lists for the communication of developers and users. Researchers have been mining the history of such lists to study communities' behavior, organization, and evolution. A potential threat of this kind of study is that users often use multiple email addresses...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Institutional Knowledge at Singapore Management University
2016
|
| Subjects: | |
| Online Access: | https://ink.library.smu.edu.sg/sis_research/8774 https://ink.library.smu.edu.sg/context/sis_research/article/9777/viewcontent/icsme16b.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Singapore Management University |
| Language: | English |
| Summary: | Many software projects adopt mailing lists for the communication of developers and users. Researchers have been mining the history of such lists to study communities' behavior, organization, and evolution. A potential threat of this kind of study is that users often use multiple email addresses to interact in a single mailing list. This can affect the results and tools, when, for example, extracting social networks. This issue is particularly relevant for popular and long-term Open Source Software (OSS) projects, which attract participation of thousands of people. Researchers have proposed heuristics to identify multiple email addresses from the same participant, however there are few studies analyzing the effectiveness of these heuristics. In addition, many studies still do not use any heuristics for authors' disambiguation, which can compromise the results. In this paper, we compare six heuristics from the literature using data from 150 mailing lists from Apache Software Foundation projects. We found that the heuristics proposed by Oliva et al. and a Naïve heuristic outperformed the others in most cases, when considering the F-measure metric. We also found that the time window and the size of the dataset influence the effectiveness of each heuristic. These results may help researchers and tool developers to choose the most appropriate heuristic to use, besides highlighting the necessity of dealing with identity disambiguation, mainly in open source software communities with a large number of participants. |
|---|