Linguistic Analysis of Crowd Requirements: An Experimental Study

Abstract

Users of today's online software services are often diversified and distributed, whose needs are hard to elicit using conventional RE approaches. As a consequence, crowd-based, data intensive requirements engineering approaches are considered important. In this paper, we have conducted an experimental study on a dataset of 2,966 requirements statements to evaluate the performance of three text clustering algorithms. The purpose of the study is to aggregate similar requirement statements suggested by the crowd users, and also to identify domain objects and operations, as well as required features from the given requirements statements dataset. The experimental results are then cross-checked with original tags provided by data providers for validation.