Requirements Elicitation (RE) is focused on identifying and characterizing the stakeholders and their requirements. Such an activity may be challenging as the scope of the software product domain grows, generating errors and delays. Natural Language Processing (NLP) deals with automatically analyzing, understanding, and generating natural language. Software analysts use NLP-based approaches for improving RE, making it more efficient and reliable. However, domain scope and limitation for understanding the writing styles of requirements documents generate significant drawbacks for such approaches. In this Ph.D. Thesis we propose SQUARE (Scalable QUestion Answering for Requirements Elicitation), a novel approach for improving the NLP-based approaches for RE based on Question Answering Systems (QASs), comprising a meta-restricted domain for RE and a rule-based approach for generating RE-related questions and answers. QASs are used for extracting precise and concise answers to natural language questions. The SQUARE model represents a contribution for the NLP-based approaches for RE, allowing software analysts for identifying, extracting, and structuring key abstractions from requirements documents such as actors, actions, and concepts in a more natural way due to its proximity to a real-life RE domain. We validate our proposal by using an experimental process. The SQUARE model is included as a new work product for eliciting requirements. Therefore, the SQUARE model is intended to be an NLP-based approach to RE for software analysts.