(Semi-)automatic requirements categorization

Paper on (semi-)automatic requirements categorization accepted by REFSQ: In this joint work with Daniel Ott  we investigated how to support a team of engineers in categorizing requirements efficiently and consistently based on a socio-technical approach. In the paper, we investigate levels of automation and their effect on organizational learning.

Knauss, Eric and Ott, Daniel: (Semi-)Automatic Categorization of Natural Language Requirements. 20th International Working Conference on Requirements Engineering: Foundation for Software Quality (REFSQ’ 14), Essen, Germany, 2014


Context and motivation: Requirements of todays industry specifications need to be categorized for multiple reasons, including analysis of certain requirement types (like non-functional requirements) and identification of dependencies among requirements.This is a pre-requisite for effective communication and prioritization of requirements in industry-size specifications.

Question/problem: Because of the size and complexity of these specifications, categorization tasks must be specifically supported in order to minimize manual efforts and to ensure a high classification accuracy. Approaches that make use of (supervised) automatic classification algorithms have to deal with the problem to provide enough training data with excellent quality.

Principal ideas/results: In this paper, we discuss the requirements engineering team and their requirements management tool as a socio-technical system that allows consistent classification of requirements with a focus on organizational learning. We compare a manual, a semi-automatic, and a fully-automatic approach for the classification of requirements in this environment. We evaluate performance of these approaches by measuring effort and accuracy of automatic classification recommendations and combined performance of user and tool, and capturing the opinion of the expert-participants in a questionnaire. Our results show that a semi-automatic approach is most promising, as it offers the best ratio of quality and effort and the best learning performance.

Contribution: Our contribution is the definition of a socio-technical system for requirements classification and its evaluation in an industrial setting at Mercedes-Benz with a team of ten practitioners.


requirements, classification, categorization, natural language


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s