Deteksi Konflik pada Spesifikasi Kebutuhan Perangkat Lunak Berbasis Aturan dan Similaritas Wordnet

Sarwosri, Sarwosri (2025) Deteksi Konflik pada Spesifikasi Kebutuhan Perangkat Lunak Berbasis Aturan dan Similaritas Wordnet. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

Text 7025201002-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (2MB) | Request a copy

Abstract

Tahapan spesifikasi kebutuhan merupakan tahap awal dalam siklus hidup pengembangan perangkat lunak dan memiliki peran krusial karena memengaruhi keberhasilan atau kegagalan tahap-tahap berikutnya, terutama perancangan. Salah satu elemen dalam dokumen spesifikasi kebutuhan adalah pernyataan kebutuhan fungsional, yang ditulis dalam bahasa alami. Namun, penggunaan bahasa alami dapat menyebabkan ambiguitas dan konflik, terutama konflik yang disebabkan oleh inkonsistensi antar kalimat kebutuhan. Konflik semacam ini dapat muncul pada berbagai aspek kalimat, seperti operasi, input, output, dan batasan.
Penelitian sebelumnya menggunakan metode clustering untuk mengelompokkan kebutuhan yang konflik, namun hasilnya belum dapat langsung digunakan oleh pengembang sistem karena masih memerlukan pemeriksaan manual terhadap konflik antar kalimat dalam satu cluster. Selain clustering, pendekatan berbasis aturan juga telah digunakan untuk mendeteksi konflik, seperti yang dilakukan oleh Chentouf. Sistem berbasis aturan yang dikembangkan Chentouf mendeteksi konflik menggunakan tipe-tipe konflik yang terdefinisi dengan baik dan aturan yang terstruktur, namun belum mengakomodasi tuple batasan, sinonim, dan antonim. Penelitian ini berupaya melengkapi kekurangan tersebut dengan dua kontribusi utama: (1) menghasilkan keluaran berupa pasangan kalimat kebutuhan, jenis konflik, dan bagian tuple yang konflik, serta (2) menambahkan analisis berbasis tuple batasan, sinonim, dan antonim.
Metodologi penelitian ini mencakup empat tahapan utama: perbaikan dataset, pemrosesan kalimat, deteksi konflik, dan evaluasi performa sistem. Tahap penyiapan dataset melibatkan pengumpulan dan modifikasi dataset agar menjadi atomik. Pemrosesan kalimat mencakup preprocessing kata, penandaan Part-of-Speech (POS), dan parsing dependensi, yang hasilnya digunakan untuk anotasi. Anotasi mencakup pelabelan delapan tuple dalam kalimat kebutuhan format EARS <even> <Id> <kelompokId> <agen> <operasi> <input> <output> batasan>, format User Story <Id> <kelompokId> <agen> operasi> <obyek> <tujuan>. Konflik antar kalimat kebutuhan dideteksi menggunakan sistem berbasis aturan dan WordNet::Similarities, dengan hasil berupa pasangan kalimat konflik, jenis konflik, dan bagian tuple yang teridentifikasi.
Evaluasi dilakukan pada lima dataset publik EARS (UAV, UAVR, UAV+ WorldVista, OpenCoss) serta dataset user story (g02, g10, g17, g27, g27-R). Hasil menunjukkan bahwa metode berbasis aturan yang dikombinasikan dengan Dependency Parsing dan WordNet pada EARS meningkatkan akurasi deteksi sebesar 8,55% , dan metode berbasis aturan yang dikombinasikan dengan WordNet pada User Story sebesar 29,65% dibandingkan metode berbasis aturan saja.

====================================================================================================================================

The requirements specification phase serves as the initial stage in the software development life cycle and plays a crucial role in determining the success or failure of subsequent stages, particularly design. One of the key elements in the requirements specification document is the functional requirements statement, which is typically written in natural language. However, the use of natural language can lead to ambiguities and conflicts, particularly those arising from inconsistencies between requirement statements. Such conflicts may occur across various aspects of the statements, including operasis, inputs, outputs, and constraints.
Previous studies have employed clustering methods to kelompok conflicting requirements. Still, their outcomes have not been directly usable by system developers, as manual inspection of conflicts within a cluster is still required. In addition to clustering, rule-based approaches have also been utilized for conflict detection, as demonstrated by Chentouf. Chentouf’s rule-based system detects conflicts using well-defined conflict types and structured rules. However, it does not accommodate constraint tuples, synonyms, and antonyms. This study aims to address these limitations through two primary contributions: (1) producing outputs in the form of requirement sentence pairs, conflict types, and conflicting tuple components; and (2) incorporating tuple-based analysis of constraints, synonyms, and antonyms.
The research methodology consists of four main stages: dataset refinement, sentence processing, conflict detection, and system performance evaluation. The dataset preparation stage involves collecting and modifying datasets to ensure atomicity. Sentence processing includes word preprocessing, Part-of-Speech (POS) tagging, and dependency parsing, with the results used for annotation. Annotation involves labeling eight tuples within the requirement sentences EARS format <event> <Id> <groupId> <agent> <operation> <input> <output> <contraint>, User Story format <Id> <groupId> <agent> <operation> <object> <goal>. Conflicts between requirement sentences are detected using a rule-based system and WordNet::Similarities, resulting in outputs comprising conflicting sentence pairs, conflict types, and identified conflicting tuple components.
The evaluation was conducted on five public EARS datasets (UAV, UAV-R, UAV+, WorldVista, OpenCoss) and user story datasets (g02, g10, g17, g27, g27-R). The results indicate that the rule-based method combined with Dependency Parsing and WordNet for EARS improves detection accuracy by 8.55%, and The Rule-based method and WordNet for User Story by 29.65% compared to the rule-based method alone.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Kebutuhan Fungsional, Spesifikasi Kebutuhan Perangkat Lunak, EARS, User Story, Konflik, Rule-Based, WordNet. Functional Requirements, Software Requirements Specification, EARS, User Story, Conflict, Rule-Based, WordNet.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55001-(S3) PhD Thesis (Comp Science)
Depositing User:	Sarwosri Sarwosri
Date Deposited:	02 Feb 2025 07:30
Last Modified:	02 Feb 2025 07:30
URI:	http://repository.its.ac.id/id/eprint/117801

Actions (login required)

View Item