The advantages of controlled vocabulary in numbers
Word has long since gotten around that in systems engineering we should not just start writing. Instead, we should control the language. The vocabulary to be used is an important aspect that can be easily controlled.
I recently came across a study that deals with exactly this topic, the effects of controlled vocabulary in requirements management. More importantly, the effects are quantified. In the following, I will first answer the question of what forms of vocabulary there are. I will then list a few interesting figures from the study that quantify the added value of controlled vocabulary.
The use of controlled vocabulary (CVs) aims to increase the quality of the specifications of the software requirements. So far so good: the quality of the written documentation is better. Controlled vocabulary also helps to avoid ambiguities and Complexity to reduce.
The study
The authors of the study were surprised that very little had been published on the effects of controlled vocabulary. They therefore carried out a meta-study in which they examined 90 papers in detail from 2348 published papers on the topic. The paper published in 2020 can be downloaded here:
What is controlled vocabulary?
The most commonly known controlled vocabulary is probably the Glossary. This is not even mentioned in the paper. Instead, the authors only identify the two groups Ontology and taxonomy.
At first glance, this seems somewhat simplistic. And in reality, the authors have also identified two other groups of controlled vocabulary: Folksonomies and thesauri.
Folksonomies are also referred to as "social tags". The authors have found papers on this, but not in the context of requirements management.
Thesaurior synonym dictionaries, appear in many papers, but nowhere as the main topic. All the papers examined discussed thesauri in the context of ontologies,
The glossary is not mentioned at all, as it is a simplified ontology.
What activities does the controlled vocabulary support?
The authors identified five activities in requirements management in which controlled vocabulary plays a relevant role. The list should surprise few readers:
All papers emphasize the need for clear, objective and unambiguous writing. The recording in written form occurs during specification and elicitation, therefore the use of controlled vocabulary is strongest here.
However, controlled language is also helpful in the analysis phase. Using it makes it much easier to recognize conflicts and contradictions, for example.
Which aspects of the development process and the product are influenced by the use of controlled language?
This is actually the most exciting question. Because it shows how the controlled vocabulary can provide real added value. Here are the results first:
Controlled vocabulary plays by far the biggest (positive) role in leading the team and creating a common understanding. This in turn is the basis for communication, which is listed much further down with 5 sources.
The number of sources that see the controlled vocabulary as the basis for automation and tool support is surprisingly high. More on this in a moment.
In my opinion, many of the aspects fall into the area of quality. If you were to combine completeness, accuracy, ambiguity and consistency with "quality", we would get 35 mentions, which would also put us in first place.
The topic of variants and product lines is also becoming increasingly important. It is therefore not surprising that as many as 8 sources mentioned reusability.
Too bad, but not surprisingly, only one source opens Modeling a
Automation
For some time now, I have been working on the topic of automation in product development with artificial intelligence (AI) apart. The result of this is the virtual quality assistant Semiantwhich is already in use in an industrial environment. In this respect, the results of this study confirm that there is enormous potential in this area.
Ontologies in particular often provide the theoretical foundation for automation. This includes tasks such as detecting conflicts, classifying requirements or recognizing errors. Various tools are already available, particularly in the area of non-functional requirements. The processing of natural language is drastically simplified and much more accurate thanks to the controlled vocabulary. Some tools can also semi-automatically convert texts into requirements based on Text templates based.
Further results
I recommend that academic readers in particular at least skim the paper. The authors also discuss the scientific environment: Who are the key people doing research in this area? In which countries is research being carried out? Great Britain is in first place, Germany in ninth place. Of course, such statistics are only of limited value, as much industrial research is not published. The best-placed company is Airbus (UK).
What happens next?
As is usual with such papers, the authors have included a list of their research objectives. I will follow the authors' work closely.
Of course, I am even more interested in the practical application. That's why I'm going to take a closer look at the literature on automation in order to explore its potential for integration in Semiant.
Semiant already has an AI-supported glossary creation that is used for various automation scenarios. Interested? Then please contact Subscribe to Semiant mailing list >>
