Derive the contextual information from developer’s questions

Introduction

It’s natural for developers to ask questions about software systems when they perform various tasks like refactoring, searching, developing a new feature and understanding the source code. Common questions are asked regarding classes, functions, and project structure etc. Understanding a question fully can be very useful to derive the answer.

For example: Question: What are the deprecated classes? This question is about classes, developers want to retrieve classes that are annotated as deprecated. This question can be asked in a different domain, for example, education, development. In development also, it can be asked in different activities like refactoring, dead code elimination, testing etc. and also who asked this question, can also play an important role to predict next information about this question.

Extracting this contextual information from questions could significantly improve the quality of answers, due to contextual knowledge.

Problem

However, this approach suffers heavily from ambiguous words, grammar and missing information from a questioner. Therefore, the extraction of this information can be very challenging.

Aim

This project will shed light on extracting contextual information of a question, and finding potential development activities or topics related to the question.

Steps

  • Study of various features that support extraction of contextual information from developer questions written in natural language.
  • Extraction of subjects (nouns), actions (verbs), and their relations.

  • Establishment of an approach to identify the topic of a question. The topics are restricted to five main domains that are testing, refactoring, user-interface, coding style, and general analysis.
  • Experimentation with topic modeling using different approaches to explore potential benefits.
  • Evaluation of the results.

Contact

Pooja Rani

Last changed by pooja on 18 September 2018