Evaluating recommended applications

  • Authors:
  • Mark Grechanik;Denys Poshyvanyk

  • Affiliations:
  • Accenture Technology Labs, Chicago, IL;The College of William and Mary, Williamsburg, VA

  • Venue:
  • Proceedings of the 2008 international workshop on Recommendation systems for software engineering
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Large open source software repositories are polluted with incomplete or inadequately functioning projects having scarce or poor descriptions. Developers often search these repositories to find sample applications containing implementations of relevant features. While relying on software search engines that retrieve germane applications based on direct matches between user queries and words in the descriptions (or source code files), it is difficult to warrant that retrieved applications contain functionality described by their authors in project summaries. We propose a novel approach called K9 that helps users evaluate if recommended applications implement functionality, which is described by the authors in project summaries. Using programmers' queries describing high-level concepts (e.g., sql server, floating toolbar, or smart card), K9 uses information retrieval and program analysis techniques to evaluate recommended applications based on how they implement these high-level concepts. We conjecture that K9 will effectively support developers in evaluating functionality of the retrieved applications.