Supporting Query Formulation for Text Retrieval Applications in Software Engineering

Text Retrieval (TR) techniques have been successfully used to leverage the textual information found in software artifacts with the purpose of aiding developers with their daily tasks. TR techniques require a query as input and the usefulness of the results they retrieve depends greatly on this query. While some queries retrieve relevant information for the current task, others do not, therefore pointing developers in the wrong direction. Developers have a hard time realizing this before going through the search results, which, in the case of "bad" queries means time and effort lost looking at irrelevant information. In this scenario, developers have to reformulate the query, often without pointers on how to improve it. The work presented in this paper introduces novel approaches to address these challenges and makes two main contributions: 1) defines the first approach for predicting the success of a TR query in the context of SE tasks, 2) introduces automatic approaches that analyze a query and improve it by finding the most suited reformulation for it. The approaches were evaluated for the task of concept location in source code and the results of the performed studies reveal their usefulness.