This study is aimed to validate a text processing architecture for knowledge acquisition and analyze the performance of several populations under controlled validation studies by focusing on empirical methods. We report our experience by analyzing three case studies whose performance issues have been addressed by using a proposed natural-language-mapping model. We discuss the advantages and the disadvantages of an automated prototype for implementing such architectural model, compared with a by-hand one. We compare the cases running out on the prototype in order to identify the suitable features the software systems should have. The final goal of the validation process is describing an ongoing research work concerned with the definition of an approach to automate processes for knowledge extraction in requirements engineering. Some of achieved findings of the performance analysis are: (i) the approach can be applied to business-based technical documents regardless of the organizational process involved; (ii) the activities related to the domain understanding can be executed in a low-costs process; and (iii) empirical methods can be used in controlled validation studies for knowledge extraction approaches.