Automated analysis of textual use-cases: Does NLP components and pipelines matter?

Abstract

Significant time is spent by practitioners to analyze use-cases written in Natural Language (NL). With only prescriptive templates to describe complex scenarios, common errors like misinterpretation and oversight can have costly consequence later during system development. A semi-automatic approach based on NL processing can reduce the time spent on requirement analysis and bootstrap design activity. However, linguistic community has adopted pipeline processing to handle NL ambiguities where several sequential tasks aid in solving a bigger task. Choosing NL processing techniques depends on the domain and task to accomplish. As use-cases are domain specific it is crucial to identify suitable pipelines to process them. This is highlighted in our evaluation of two pipelines consisting of syntactic and semantic techniques on use-cases found in theory and practice. We believe, the promising results has opened up the need for exploring more task specific NLP pipelines and evaluation thereof. © 2012 IEEE.