Mining Twitter Feeds for Software User Requirements

Document Type

Conference Proceeding

Publication Date

9-22-2017

Abstract

Twitter enables large populations of end-users of software to publicly share their experiences and concerns about software systems in the form of micro-blogs. Such data can be collected and classified to help software developers infer users' needs, detect bugs in their code, and plan for future releases of their systems. However, automatically capturing, classifying, and presenting useful tweets is not a trivial task. Challenges stem from the scale of the data available, its unique format, diverse nature, and high percentage of irrelevant information and spam. Motivated by these challenges, this paper reports on a three-fold study that is aimed at leveraging Twitter as a main source of software user requirements. The main objective is to enable a responsive, interactive, and adaptive data-driven requirements engineering process. Our analysis is conducted using 4,000 tweets collected from the Twitter feeds of 10 software systems sampled from a broad range of application domains. The results reveal that around 50% of collected tweets contain useful technical information. The results also show that text classifiers such as Support Vector Machines and Naive Bayes can be very effective in capturing and categorizing technically informative tweets. Additionally, the paper describes and evaluates multiple summarization strategies for generating meaningful summaries of informative software-relevant tweets.

Publication Source (Journal or Book title)

Proceedings - 2017 IEEE 25th International Requirements Engineering Conference, RE 2017

First Page

1

Last Page

10

This document is currently not available here.

Share

COinS