Article Preview
Top1. Introduction
Modern methods in information technology enable software engineers to develop various software quickly and effectively with different styles. In this paper, the problem addressed is related to the requirement analysis in the software development phase. Recent trends of software engineering largely depend on object-oriented design paradigm that uses the Unified Modeling Languages (UML). UML is used for modeling the user software requirements, documenting the software assets, software development and redevelopment (Perez-Gonzalez, 2002). The UML class model is the core for object-oriented analysis and design. The existing tools such as ReBuilder (Oliveria et al., 2006), CM-Builder (Harmain & Gaizauskas, 2003), GOOAL (Perez-Gonzalez & Kalita, 2002), NL-OOML (Anandha & Uma, 2006) and UML-Generator (Bajwa et al., 2009) attempt to generate UML class diagrams automatically from the natural languages.
The problem with these tools is that they generate the class diagram with lower accuracy due to the informal nature of NL and its ambiguity (Li et al., 2005; Mich, 2001). This paper is focused on designing the perfect class models from the user requirements. The user specifies their requirements in natural languages such as English. Since natural language processing is a difficult task this research work is divided into two phases. The initial phase is parsing and in this phase, we make use of OpenNLP (Mohanan et al., 2016). The second phase is a transformation phase and for transformation phase, Semantics of Business Vocabulary and Rules (SBVR) (Feuto et al., 2013) process is implemented.
The OpenNLP is used to produce the POS (parts of speech) (Toutanova & Manning, 2000) tags of SRS which is in the form of a natural language such as English. The POS tags contain the required details such as noun, verb, adverb, etc., of the sentences of SRS. We use Apache OpenNLP to implement the pre-processing phase and to carry out Sentence splitting, tokenization and POS tagging. In the second phase Semantics of Business Vocabulary and Rules process generates the vocabulary and rules. Since SBVR has natural language syntax it provides a suitable way to capture the Object-Oriented items in requirement specifications and also easy to understand by both the user and machine (Fernandez et al., 2000; Lane & Henderson, 2001; Bajwa et al., 2011). The SBVR vocabulary extraction, SBVR rules generation, Object Oriented Analysis of SBVR rules and UML class model generation are the phases in SBVR process. The remainder of this paper is organized as below.
Section 2 describes some related works which are available in the literature. Section 3 describes the class identification method, its preliminaries, explains the SRS to class identification using SBVR, object oriented analysis from the SBVR rule and class diagram generation. Section 4 explains the evaluation of our methodology followed in this paper and includes experimental results of various case studies of this work which are used to demonstrate precision and recall of the proposed concept. Section 5 concludes this paper.