Please use this identifier to cite or link to this item: http://cmuir.cmu.ac.th/jspui/handle/6653943832/69378
Title: Alternative Elicitation Framework to Improve Requirement Quality for Very Small Enterprise Using Knowledge Management Approach
Other Titles: กรอบความคิดการสกัดข้อมูลทางเลือกเพื่อพัฒนาคุณภาพความต้องการสำหรับผู้ประกอบการขนาดเล็กมากโดยใช้วิธีการจัดการความรู้
Authors: Pornpen Lertthasanawong
Authors: Lect.Dr.Tirapot Chandarasupasang
Asst.Prof.Dr.Nopasit Chakpitak
Asst.Prof.Dr.Pitipong Yodmongkol
Pornpen Lertthasanawong
Issue Date: Nov-2014
Publisher: เชียงใหม่ : บัณฑิตวิทยาลัย มหาวิทยาลัยเชียงใหม่
Abstract: Most of software companies in Thailand are categorized as very small enterprises (VSEs) which have many limitations such as resource, fund and experience. The critical problem of the software industry is to determine the scope and understand the customer requirements. Requirement Elicitation is the first process of gathering customer requirements. So this process is very significant for software development. Most of the customer software requirements are expressed in natural language. Some language has a specific character for example Chinese, Japanese, and Thai languages. They are written in un-segmented form and have no punctuation mark. This causes the reader to ambiguity and mis-interpretation. From the survey, it is found that the problems occured during the communication between the customers and the software companies. The customers do not know what exactly they want and the software companies have no tool and technique for deeper understanding. The software companies are depend on the experience of system analysts who may have no experience in customer business. This research is developed the elicitation tool. The conceptual frame work is based on CommonKADS which is to formulate knowledge models to manage knowledge creation and reuse. The research used Natural Language Processing for vector representation and then group all requirement representation by keywords. This method can reduce the duration of the work, reduce the cost of software development and reduce resources which are the limitations of VSEs. The research methodology is called TDVC (Transaction Domain Dictionary Vector Space Representation and Cluctering) There are four main stages of the research approaches which are (1) Transaction Dialog construction. At the end of this stage all requirements are collected. (2) Domain Definition. Each software requirement is tokenized into a series of terms before it can be further analyzed and translated. The unknown words from software requirements are trained for the new vocabularies and collected for specific business dictionary meaning which is called Domain Dictionary. (3) Neo Vector Space representation. the transformative of requirement into vector space for each requirement sentence. (4) New grouping. The keywords similarity of inner group and outer group is compared for suitable of requirement grouping which can be grouped correctly. In this research, there are four cases study of accounting requirements. These cases are the most developed case of VSEs. These cases are small packaged software in Thai language, small customized software in Thai language, medium customized software in Thai language and finally small packaged software in English language for robustness test of real use. The results showed that TDVC can ensure the precision (P), integral (Recall) and the F-measure in high number. The results also showed that the research methodology gives the high quality criteria in correctness, completeness, traceability and consistency when compared to the traditionl method such as Interview, Focus groups, Brainstorming, Scenarios, Storyboards, Prototyping and Joint Application Development (JAD) There are many research findings, firstly the TDNN can be used for Business Analysis in real situation. Secondly, the methodology can be implemented for special character language such as Thai and also English. Thirdly, The method can help to translate the requirements because the requirement representations are in vector from the domain expert knowledge base. This method can reduce the ambiguity which are from human decisions making. Finally, the method shows better performance than the traditional method in term of cost effectiveness of the elicitation process, quality of elicitation products and quality of elicitation services.
URI: http://cmuir.cmu.ac.th/jspui/handle/6653943832/69378
Appears in Collections:CAMT: Theses

Files in This Item:
File Description SizeFormat 
Full.pdf5.08 MBAdobe PDFView/Open    Request a copy


Items in CMUIR are protected by copyright, with all rights reserved, unless otherwise indicated.