Automated analysis of requirement specifications
ICSE '97 Proceedings of the 19th international conference on Software engineering
Exploring Requirements: Quality Before Design
Exploring Requirements: Quality Before Design
SEW '01 Proceedings of the 26th Annual NASA Goddard Software Engineering Workshop
Introduction to special issue on machine learning approaches to shallow parsing
The Journal of Machine Learning Research
A simple but useful approach to conjunct identification
ACL '92 Proceedings of the 30th annual meeting on Association for Computational Linguistics
Market research for requirements analysis using linguistic tools
Requirements Engineering
An unsupervised model for statistically determining coordinate phrase attachment
ACL '99 Proceedings of the 37th annual meeting of the Association for Computational Linguistics on Computational Linguistics
ACL '02 Proceedings of the 40th Annual Meeting on Association for Computational Linguistics
On the Systematic Analysis of Natural Language Requirements with CIRCE
Automated Software Engineering
CIT '06 Proceedings of the Sixth IEEE International Conference on Computer and Information Technology
The case for dumb requirements engineering tools
REFSQ'12 Proceedings of the 18th international conference on Requirements Engineering: foundation for software quality
Hi-index | 0.00 |
[Context and Motivation] Many a tool for finding ambiguities in natural language (NL) requirements specifications (RSs) is based on a parser and a parts-of-speech identifier, which are inherently imperfect on real NL text. Therefore, any such tool inherently has less than 100% recall. Consequently, running such a tool on a NL RS for a highly critical system does not eliminate the need for a complete manual search for ambiguity in the RS. [Question/Problem] Can an ambiguity-finding tool (AFT) be built that has 100% recall on the types of ambiguities that are in the AFT's scope such that a manual search in an RS for ambiguities outside the AFT's scope is significantly easier than a manual search of the RS for all ambiguities? [Principal Ideas/Results] This paper presents the design of a prototype AFT, SREE (Systemized Requirements Engineering Environment), whose goal is achieving a 100% recall rate for the ambiguities in its scope, even at the cost of a precision rate of less than 100%. The ambiguities that SREE searches for by lexical analysis are the ones whose keyword indicators are found in SREE's ambiguity-indicator corpus that was constructed based on studies of several industrial strength RSs. SREE was run on two of these industrial strength RSs, and the time to do a completely manual search of these RSs is compared to the time to reject the false positives in SREE's output plus the time to do a manual search of these RSs for only ambiguities not in SREE's scope. [Contribution] SREE does not achieve its goals. However, the time comparison shows that the approach to divide ambiguity finding between an AFT with 100% recall for some types of ambiguity and a manual search for only the other types of ambiguity is promising enough to justify more work to improve the implementation of the approach. Some specific improvement suggestions are offered.