An Application of Natural Japanese Processing to Software Requirement Auditing
Grant-in-Aid for Scientific Research (C)
|Research Institution||TOKAI UNIVERSITY|
KAMIJO Fumihiko Tokai University, Information Science Laboratory, Professor, 情報処理研究教育施設, 教授 (50224663)
|Project Fiscal Year
1995 – 1997
Completed(Fiscal Year 1997)
|Budget Amount *help
¥2,200,000 (Direct Cost : ¥2,200,000)
Fiscal Year 1997 : ¥400,000 (Direct Cost : ¥400,000)
Fiscal Year 1996 : ¥800,000 (Direct Cost : ¥800,000)
Fiscal Year 1995 : ¥1,000,000 (Direct Cost : ¥1,000,000)
|Keywords||Requirement Analysis / Requirement Definition / Japanese / Software / Computer Science / 要求分析 / 要求定義 / 日本語 / ソフトウェア / コンピュータ科学 / 自然語 / 要求分析・定義 / レキシコン / 自然語処理|
Requirement Specification is the first phase of a software development project in a lifecycle. A requirement specification is often incomplete and ambiguous due to a lack of detailed system information.
There are two approaches to cope with the problem. One method is to use natural language with less efforts and audit it carefully, another is to employ a formal tool to define a precise requirement with a great effort. In this research we chose Japanese language as a representative of the natural languages and examined the possibility to audit it automatically.
There are various problems to audit a requirement specification by a machine. Among them, it is difficult to separate optimal phrase structure for a closed context to examine a terminal idea. Some conventional Japanese processing schemes such as (1) morpheme analysis, (2) syntactic analysis and (3) automatic segmentation are applicable. It is necessary to recombine some of morphemes into a terminal idea.
There is a possibility to so
lve these problems by application of : (a) Japanese lexicon for computer use, e.g.IPAL(b) Standard terminology defined in ISO/IEC 12207 (SLCP), and (c) Contextual knowledge of a specific application domain.
Extensive experiments were done, using the formal tax return statements as a specimen specification. Outcomes of experiments were as follows : (i) Requirement analysis using morpheme analysis, syntactic analysis and automatic segmentation was possible, (ii) Multi-level concept-linking was possible with a locally generated database called "triplets", (iii) Conceptual structure of a requirement description might be described as "concept tree, " which show any terminal idea.
Some strategies to be tested as the means of requirement analyzes are still in the state of "future plan." They are : " Automatic construction of the "concept tree, " "Generalized SLCP terminology database, " "Analysis of the intentionally ambiguous representation, " "Automatic software tools for requirement specification auditing."
It is difficult to obtain the same result to the process of a human. There is another problem in the specification analysis, which is the intentionally built-in ambiguity in "governmental" statements.
Research Output (5results)