TU Darmstadt / ULB / TUbiblio

Evaluating the Evaluations of Code Recommender Systems: A Reality Check

Proksch, Sebastian and Amann, Sven and Nadi, Sarah and Mezini, Mira :
Evaluating the Evaluations of Code Recommender Systems: A Reality Check.
[Online-Edition: http://doi.acm.org/10.1145/2970276.2970330]
In: International Conference on Automated Software Engineering pp. 111-121. ISSN 978-1-4503-3845-5
[Article] , (2016)

Official URL: http://doi.acm.org/10.1145/2970276.2970330

Abstract

While researchers develop many new exciting code recommender systems, such as method-call completion, code-snippet completion, or code search, an accurate evaluation of such systems is always a challenge. We analyzed the current literature and found that most of the current evaluations rely on artificial queries extracted from released code, which begs the question: Do such evaluations reflect real-life usages? To answer this question, we capture 6,189 fine-grained development histories from real IDE interactions. We use them as a ground truth and extract 7,157 real queries for a specific method-call recommender system. We compare the results of such real queries with different artificial evaluation strategies and check several assumptions that are repeatedly used in research, but never empirically evaluated. We find that an evolving context that is often observed in practice has a major effect on the prediction quality of recommender systems, but is not commonly reflected in artificial evaluations.

Item Type: Article
Erschienen: 2016
Creators: Proksch, Sebastian and Amann, Sven and Nadi, Sarah and Mezini, Mira
Title: Evaluating the Evaluations of Code Recommender Systems: A Reality Check
Language: English
Abstract:

While researchers develop many new exciting code recommender systems, such as method-call completion, code-snippet completion, or code search, an accurate evaluation of such systems is always a challenge. We analyzed the current literature and found that most of the current evaluations rely on artificial queries extracted from released code, which begs the question: Do such evaluations reflect real-life usages? To answer this question, we capture 6,189 fine-grained development histories from real IDE interactions. We use them as a ground truth and extract 7,157 real queries for a specific method-call recommender system. We compare the results of such real queries with different artificial evaluation strategies and check several assumptions that are repeatedly used in research, but never empirically evaluated. We find that an evolving context that is often observed in practice has a major effect on the prediction quality of recommender systems, but is not commonly reflected in artificial evaluations.

Journal or Publication Title: International Conference on Automated Software Engineering
Publisher: IEEE/ACM
Uncontrolled Keywords: General and reference: Evaluation Information systems: Recommender systems Software and its engineering: Software notations and tools Human-centered computing: Design and evaluation methods Empirical Study Artificial Evaluation IDE Interaction Data
Divisions: Department of Computer Science
Department of Computer Science > Software Technology
Date Deposited: 28 Sep 2016 09:08
Official URL: http://doi.acm.org/10.1145/2970276.2970330
Identification Number: doi:10.1145/2970276.2970330
Related URLs:
Funders: German Federal Ministry of Education and Research (BMBF) with grant no. 01IS12054, German Science Foundation (DFG) in the context of the CROSSING Collaborative Research Center (SFB #1119, project E1)
Export:

Optionen (nur für Redakteure)

View Item View Item