Improved query reformulation for concept location using CodeRank and document structures

Abstract

Abstract-During software maintenance, developers usually deal with a significant number of software change requests. As a part of this, they often formulate an initial query from the request texts, and then attempt to map the concepts discussed in the request to relevant source code locations in the software system (a.k.a., concept location). Unfortunately, studies suggest that they often perform poorly in choosing the right search terms for a change task. In this paper, we propose a novel technique -ACERthat takes an initial query, identifies appropriate search terms from the source code using a novel term weight -CodeRank, and then suggests effective reformulation to the initial query by exploiting the source document structures, query quality analysis and machine learning. Experiments with 1,675 baseline queries from eight subject systems report that our technique can improve 71% of the baseline queries which is highly promising. Comparison with five closely related existing techniques in query reformulation not only validates our empirical findings but also demonstrates the superiority of our technique.

Improving IR-based bug localization with context-aware query reformulation

“…To the best of our knowledge, to date, graph-based term weighting has been employed only on unstructured regular texts [42] and semi-structured source code [41]. On the contrary, we deal with stack traces which are structured and should be analysed carefully.…”

“…Hence, they need to be complemented with appropriate keywords before using. A recent study [41] provides improved reformulations to a poor natural language query for concept location by first collecting pseudo-relevance feedback and then employing graph-based term weighting. In pseudo-relevance feedback, Top-K result documents, returned by a given query, are naively considered as relevant and hence, are selected for query reformulation [12,20].…”

“…Thus, the above phrase adds four distinct nodes and three connecting edges -"get"↔"context", "context"↔"class" and "class"↔"loader" -to the source term graph. Term Weighting using PageRank: Once each body of texts (e.g., stack traces, regular texts, source document) is transformed into a graph, we apply PageRank [11,31,41,42] to the graph for identifying important keywords. PageRank was originally designed for web link analysis, and it determines the reputation of a web page based on the votes or recommendations (i.e., hyperlinks) from other reputed pages on the web [11].…”