Consistency Detection in Software Requirement Specifications Using Natural Language Processing Techniques

Abstract

Understanding of human language text in written form by a computer tool is possible because of Natural language processing (NLP). NLP utilizes the principles of artificial intelligence (AI). The creation of NLP soft wares is a difficult task because computers only understand in some programming language which is precise, unambiguous and highly structured. Natural language text is obviously not precise. It can also be ambiguous. The linguistic structure of the sentences can also vary a lot due to slang, regional dialects and social context. Contradictory and inconsistent sentences in a set of requirements is known as conflicting requirements. In the Requirements Engineering phase of Software Development Life Cycle (SDLC) software requirements are gathered, analyzed, negotiated back and forth manually to come to a final requirements specification document that is free from a known problem – conflicting requirements. By automating conflict detection during requirements analysis phase, time, effort, and resources can be saved in going back and forth and checking for conflicts manually. Natural Language Processing is a way to pre-process software requirements contextually before a manual or automated model or algorithm can be applied on them. A fully automated tool was developed in python language using NLTK toolkit utilizing multiple NLP techniques to facilitate software requirement engineer in verifying the software requirements for consistency issues and conflicts. This tool generates a traceability matrix to identify which possible requirements have consistency issues