Método para la representación semi-automática de modelos conceptuales desde documentos de negocio escritos en lenguaje natural en español

Currently, the software development industry presents challenges for processing business information, specifically that contained in textual documents. In the process of software requirements elicitation, a potential source of relevant information is business documents, since they can facilitate the knowledge understanding about a domain, as well as know the evolution of a product. Despite its usefulness, requirements engineers do not always use it for their work because of time and costs involved. In this paper this problem is addressed and it is recognized through a systematic literature review, the potentiality of using Natural Language Processing (NLP) techniques to extract relevant textual information from business documents, and the utility of its representation in conceptual models. Starting from this, a semi-automatic method of extracting information from business documents written in natural language in Spanish and its representation in a conceptual model is proposed. The method is supported in a reference methodological framework for Text Analytics projects, is based on NLP techniques, and the output is represented in a class diagram. The method was evaluated through a case study with software analysts in Medellin-Colombia, taking as input telecommunications resolution documents. The evaluation allows us to conclude that the model is a satisfactory approach to solving the problem, and some lines of work are identified to generalize a solution.