Trends in Topics in Software Engineering Conferences, 1992 to 2016
George Mathew, Amritanshu Agrawal, Tim Menzies · arXiv (Cornell University) · 2016
Context: The very rapid pace of change in Software Engineering(SE) research makes it very difficult to gain an of the entire field or track trends in specific areas of SE. Objective: Develop a very rapid overview tool that can help humans understand how topics are evolving in SE. Method: Perform a topic modelling analysis using Tuned Latent Dirichlet Allocation of abstracts and titles of 15121 papers over the last 25 years. Results: (a)Eleven topics are identified as the dominant themes in modern SE. (b)Publication trends for topics in SE are not static; in fact, some of them become less prominent over time while others become very prominent indeed. (c) Clustering conferences on the topics they usually publish, it is observed that SE conferences fall into six large groups. (d) Considering that the Program Committee (PC) constitutes to less than a quarter of authors with published papers, the topics of papers published by the PC almost exactly match with the topics seen in all the published papers. (e) Successful authors in SE explore a diverse range of topics rather than focus on a niche set. Conclusion: Based on the results, recommendations are offered on how to(a)plan on an individual's research career, (b) intelligently make new conferences or merge older ones,(c) better recruit PCs & (d) encourage a broader range of topics at SE conferences.