Title:	USING SYSTEM MODELS TO RESOLVE CO-REFERENCE IN TRANSLATING NATURAL LANGUAGE REQUIREMENTS INTO CODE
Author(s):	Brit Arnon, David Harel and Michal Gordon-Kiwkowitz
ISBN:	978-989-8704-15-3
Editors:	Miguel Baptista Nunes, Pedro Isaías, Philip Powell and Boyan Bontchev
Year:	2020
Edition:	Single
Keywords:	Requirements to Code, Live Sequence Charts, Co-Reference Resolution, Controlled Natural Language
Type:	Full
First Page:	115
Last Page:	122
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	One important quest in software development is to be able to translate natural language requirements into executable code. In this paper we address the issue of co-reference resolution, extending our previous work on a controlled natural language translation to live sequence charts. While domain-specific co-reference resolvers perform adequately, and state-of-the-art deep learning-based generic resolvers are emerging, in software development, especially for critical systems, a higher level of performance is required. We present an approach that incorporates system-model-based anaphora resolution. Using an annotated corpus for software requirements with system models, we show that while our approach reaches high accuracy, it is synergistic to generic resolvers and when combined with them it reaches higher accuracy. Our method shows that incorporating system models into the pipeline of requirements to code is a promising approach.
	Go Back