Patterns of Requirement Writing : a systematic literature mapping

Rodrigo Cezario da Silva Fabiane Barreto Vavassori Benitti

Introduction
Patterns
help organizations to improve structure their knowledge is considered an effective resource to reuse the acquired knowledge in software development help in the process of internal communication

Requirements Patterns
are a solution which accelerates the process of software requirements elicitation establishes requirements with a higher quality of writing, increasing the quality of documents generated

Goal
What are the patterns that exist for writing software requirements?

A systematic mapping study allows the evidence in a domain to be identified at a high level of granularity.

Planning
Research Question
What are the patterns for writing software requirements currently proposed?

Search Strings
((pattern OR catalog OR category) AND (software requirement OR requirements engineering OR requirements elicitation)) ((padres OR catlogo OR categoria) AND (requisito de software OR engenharia de requisitos OR elicitao de requisitos))

Digital libraries
IEEEXplorer, ACM, IET Digital Library, SpringerLink, Science Direct, Google Scholar, CiteSeer Biblioteca Digital Brasileira de Teses e Dissertaes, Google Scholar, Banco de Teses e Dissertaes da UFRGS/ USP/ UFSCar/ UFRJ, Domnio Pblico, Biblioteca digital da SBC

Planning
Inclusion criteria
studies that would expose the pattern in detail allowing its use; studies that would provide evidence of the proposed pattern in real projects or academics.

Exclusion criteria
studies where the proposed patterns did not approach aspects related to the form of writing / documentation of software requirements; studies where the pattern was not displayed in detail, disallowing its use; studies that failed to show evidence for the application of the pattern, whether in academia or industry.

Conducting the review

The systematic mapping was conducted between June - August 2010 The execution of the formatted search string for each database resulted in 56 studies We identified 53 studies in international databases and 3 studies in national databases. However, some of these studies were mentioned by more than one database, resulting in 36 single studies.
Digital Libraries IEEEXplorer ACM IET Digital Library SpringerLink Science Direct Google Scholar CiteSeer Biblioteca Digital Brasileira de Teses e Dissertaes Google Scholar Banco de Teses e Dissertaes da UFRGS Domnio Publico Banco de Teses e Dissert. da USP Biblioteca Dig. de Teses e Dissert. UFSCar Biblioteca Digital de Teses e Dissert. - UFRJ Biblioteca Digital da SBC Results 9 13 2 5 1 16 7 0 2 0 0 0 1 0 0

Conducting the review

Reading the title and abstract of 36 studies

Reading the introduction and conclusion of 18 studies

4 primary studies identified

18 papers discarded for not meeting the criteria

13 papers discarded for not fulfilling the criteria and 1 without access to the complete paper

3 distinct patterns
(2 primary studies approached the same pattern)

Approach of Toro et al. (1999)

They present requirements templates that can improve requirements elicitation and expression - and two kinds of patterns:
Linguistic patterns, which are very used sentences in natural language requirements descriptions that can be parameterized and integrated into templates, and Requirements patterns, which are generic requirements templates that are found very often during the requirements elicitation process and that can be reused with some adaptation.

Approach of Withall (2007)

He presents 37 requirements patterns organized in 8 domains. The areas have patterns that aim to complete the functionalities of a specific domain.

Approach of Franch et al. (2010), Renault et al. (2009a), Renault et al. (2009b) e Mendez-Bonilla, Franch e Quer (2008)
In the study group an initial version of the catalogue of requirements patterns, the method for their use, called PABRE, and a meta-model for software requirement patterns were presented.
A meta-model for Software Requirement Patterns around three main concepts: 1) the structure of Software Requirement Patterns themselves; 2) the relationships among them; 3) the classification criteria for grouping them.

Comparative
Templates Toro et al. (1999) They present requirements templates and patterns that can help requirements engineers and users to elicit, express and record information systems requirements using natural language. The template provides fields for management. Patterns proposed from the analysis of two other studies (other authors). Withall (2007) The pattern is described from the definition of 9 fields. The templates presented only involve the writing of the requirement, without providing further information. Franch et al. (2010) The template only involves the writing requirement, without providing fields for additional information.

Basis for construction of pattern Relationships between patterns Categories of patterns Metamodel Tool Use / Evaluation

Patterns identified from the needs for assessment in real projects.

Does not address.

Addresses.

Generalized patterns from the analysis of 7 documents specifying requirements for completed projects (Average of 70 non-functional requirements in each document). Addresses, but not explicitly.

- Functional requirements (CRUD). - Non-functional requirements (without determining categories). Does not present There is a prototype described. More than 40 academic projects and two real projects.

- Functional requirements - Non-functional requirements (with categories). Does not present A tool is described in the author's website (1) Presents examples of more than 400 requirements using patterns.

- Non-functional requirements (extension of ISO 9126-1).

Presents The development of a tool is proposed as a future step. Evaluation in two real projects.

(1) http://www.withallyourequire.com/reqtssoftware.html

Evaluation Biases / Threats

We discarded a study for not having access to the complete paper. The conduct of the mapping was performed by only one researcher, and the stages of planning and discussion involved two researchers. The patterns presented are the result of a specific search string in a set of 15 data sources limited to the search for terms in English and Portuguese. The extension of terms, language and data sources can reveal new patterns.

Conclusions

The systematic mapping allowed to respond the research question: What are the patterns for writing software requirements currently proposed?
3 distinct patterns: Toro et al. (1999), Withall (2007) e Franch et al. (2010)

Besides the identification of patterns, we can highlight some relevant aspects:

the use of templates to guide the writing of the requirements is an option agreement in the literature; some authors prioritize the use of patterns for non-functional requirements, however, the mapping has proved to be viable to use patterns for functional requirements and non-functional requirements. At this point, two aspects should be highlighted: when the pattern applies to non-functional requirements it is recommended to do by category; and the use of patterns for functional requirements typically takes the scope of information systems, reducing the suitability for other system types .

It should be considered the relationship between the patterns, allowing patterns composed by other patterns.

Conclusions
The development of a tool that supports the use of patterns appears to be a natural way for its adoption.
We did not find citation about the use of requirement patterns by traditional tools (consolidated or well known in terms of area support requirements).

None of the studies presented a formal assessment of patterns, ie, the benefits attributed to the use of patterns resulting from an informal evaluation and more to the expression of ad-hoc review.
This finding presents an opportunity for new research.

Questions?
Rodrigo Cezario da Silva [email protected]