Abstract: Context
The automatic extraction of actors and actions (i.e., use cases) of a system from natural language-based requirement descriptions, is considered a common problem in requirements analysis. Numerous techniques have been used to resolve this problem. Examples include rule-based (e.g., inference), keywords, query (e.g., bi-grams), library maintenance, semantic business vocabularies, and rules. The question remains: can a combination of natural language processing (NLP) and artificial neural networks (ANNs) perform this job successfully and effectively?

Objective
This paper proposes a new approach to automatically identify actors and actions in a natural language-based requirements description of a system. Included are descriptions of how NLP plays an important role in extracting actors and actions, and how ANNs can be used to provide definitive identification.

Method
We used an NLP parser with a general architecture for text engineering, producing lexicons, syntaxes, and semantic analyses. An ANN was developed using five different use cases, producing different results due to their complexity and linguistic formation.

Results
Binomial classification accuracy techniques were used to evaluate the effectiveness of this approach. Based on the five use cases, the results were 17–63% for precision, 5–6100% for recall, and 29–71% for F-measure.

Conclusion
We successfully used a combination of NLP and ANN artificial intelligence techniques to reveal specific domain semantics found in a software requirements specification. An Intelligent Technique for Requirements Engineering (IT4RE) was developed to provide a semi-automated approach, classified as Intelligent Computer Aided Software Engineering (I-CASE).