Automatic development of requirement linking matrix based on semantic similarity for robust software development

Abstrak— Di dalam pengembangan perangkat lunak berskala besar, terdapat jumlah dokumen kebutuhan perangkat lunak yang sangat banyak dalam sekali proses elisitasi yang mungkin dihasilkan untuk domain yang berbeda dari setiap tim pengembang. Dokumen-dokumen ini mungkin digunakan kembali untuk mengurangi biaya dan waktu guna pengembangan perangkat lunak berikutnya. Oleh karena itu dibutuhkan suatu mekanisme untuk mendapatkan dokumen tersebut kembali yang sesuai dengan kebutuhan pengguna secara efektif dan efisien.Makalah ini mengusulkan suatu pendekatan kesamaan semantik dan struktur dalam deskripsi kasus penggunaaan guna mendapatkan kembali dokumen spesifikasi kebutuhan perangkat lunak. Dari kasus penggunaan yang didapatkan kembali tersebut kemudian dibuat suatu urutan atau ranking kesamaan kasus penggunaan berdasar suatu threshold yang telah didapatkan berdasar nilai kesepakatan tertinggi dengan pakar pada proses pengujian.Pendekatan menggunakan kesamaan semantik pada kata dan kalimat ini merupakan pendekatan yang diajukan sebagai pengganti dari pendekatan sebelumnya yaitu menggunakan term frequency dan keyword atau kata kunci. Metode ini diujicobakan pada percobaan dengan menggunakan 20 kueri deskripsi kasus penggunaan. Tiga skenario yang berbeda disusun untuk menginvestigasi nilai threshold yang ideal, mengetahui perbedaan hasil atau result set dengan pendekatan sebelumnya dan untuk mengetahui efek kombinasi struktur deskripsi kasus penggunaan pada masukan kueri terhadap hasil kueri yang didapatkan. Kata Kunci—Temu Kembali Kasus Penggunaan, Kesamaan Semantik, Sistem Temu Kembali InformasiAbstract— In software industry, there are tremendous number of software requirements documents in effect of large scale of software development. These collections of software requirements documents can be reused in order to cut off the development time and reduce the cost. There is a need to retrieve one or more of those documents which is suitable with the user’s specifications for the new software development. So, if we can retrieve many similar software requirements to a new project, development process will be less costly and less error because the retrieved software requirements can be tailored to the new case with fewer modifications. This paper presents a methodology to retrieve software requirements documents using structured base of use case description semantic similarity computation. Each element of query use case description will be calculated with the use case description in the collections or repository. The semantic similarity score is used to rank the use case in the collections and to retrieve the requirement documents to be used in the new software project development. We introduce a semantic similarity computation approach as a substitute of term frequency and keyword approach. We validate the usefulness of our method through the experiment using 20 cases. Three experiment scenarios are presented to investigate the ideal threshold value, the retrieval result differences with previous approach and to find out the effect of various combinations of structural query to the retrieval result.Keywords— Use Case Retrieval, Semantic Similarity, Information Retrieval

Download Full-text

Contextual Similarity Among Identifier Names: An Empirical Study

10.5753/vem.2020.14528 ◽

2020 ◽

Author(s):

Remo De Oliveira Gresta ◽

Elder Cirilo

Keyword(s):

Empirical Study ◽

Software Development ◽

Open Source ◽

Semantic Similarity ◽

Large Scale ◽

Current Knowledge ◽

Source Code ◽

Word Embedding ◽

Acceptable Level ◽

Domain Information

Identifiers are one of the most important sources of domain information in software development. Therefore, it is recognized that the proper use of names directly impacts the code's comprehensibility, maintainability, and quality. Our goal in this work is to expand the current knowledge about names by considering not only their quality but also their contextual similarity. To achieve that, we extracted names of four large scale open-source projects written in Java. Then, we computed the semantic similarity between classes and their attributes/variables using Fasttext, an word embedding algorithm. As a result, we could observe that source code, in general, preserve an acceptable level of contextual similarity, developers avoid to use names out of the default dictionary (e.g., domain), and files with more changes and maintained by distinct contributors tend to have better a contextual similarity.

Download Full-text

MultiCoS — A Requirements Engineering Tool

International Journal of Software Engineering and Knowledge Engineering ◽

10.1142/s021819401850002x ◽

2018 ◽

Vol 28 (01) ◽

pp. 37-55

Author(s):

Călin Eugen Nicolae Gal-Chiş ◽

Bazil Pârv

Keyword(s):

Software Development ◽

Programming Languages ◽

Semantic Similarity ◽

Web Applications ◽

Separation Of Concerns ◽

Similarity Coefficients ◽

Novel Approach ◽

Dynamic Software ◽

Multiple Skills ◽

The Web

Web development is one of the most dynamic software development branches as the web applications are popular. As applications in this domain are developed in a variety of programming languages and programmers with multiple skills in the domain are needed, also, due to various existing architectures, frameworks, and development methodologies, there are not many solutions to support properly the need for reusing resources, code or other different artefacts. The tool introduced in this paper supports MultiCoS, a novel approach to reuse requirements, resources, and other artefacts, based on the multi-dimensional separation of concerns and semantic similarity. The tool produces similarity coefficients for requirements in different applications and maps the requirements of the applications under development to reusable existing code modules based on these similarity coefficients.

Download Full-text