Resolving Ambiguities in Natural Language Software Requirements

Abstract

Requirements Engineering is one of the most vital activities in the entire Software Development Life Cycle. The success of the software is largely dependent on how well the users' requirements have been understood and converted into appropriate functionalities in the software. Typically, the users convey their requirements in natural language statements that initially appear easy to state. However, being stated in natural language, the statement of requirements often tends to suffer from misinterpretations and imprecise inferences. As a result, the requirements specified thus, may lead to ambiguities in the software specifications. One can indeed find numerous approaches that deal with ensuring precise requirement specifications. Naturally, an obvious approach to deal with ambiguities in natural language software specifications is to eliminate ambiguities altogether i.e. to use formal specifications. However, the formal methods have been observed to be cost-effective largely for the development of mission-critical software. Due to the technical sophistication required, these are yet to be accepted in the mainstream. Hence, the other alternative is to let the ambiguities exist in the natural language requirements but deal with the same using proven techniques viz. using approaches based on machine learning, knowledge and ontology to resolve them. One can indeed find numerous automated and semi-automated tools to resolve specific types of natural language software requirement ambiguities. However, to the best of our knowledge there is no published literature that attempts to compare and contrast the prevalent approaches to deal with ambiguities in natural language software requirements. Hence, in this paper, we attempt to survey and analyze the prevalent approaches that attempt to resolve ambiguities in natural language software requirements. We focus on presenting a state-of-the-art survey of the currently available tools for ambiguity resolution. The objective of this paper is to disseminate, dissect and analyze the research work published in the area, identify metrics for a comparative evaluation and eventually do the same. At the end, we identify open research issues with an aim to spark new interests and developments in this field.