Darijus' PhD Thesis page
Public defence
Thesis defended: 9th of March, 2006
Presentation: PDF
Evaluation Committee:
Prof. Pericles Loucopoulos (University of Manchester, UK)
Prof. Benkt Wangler (University of Skovde, Sweden)
Prof. Maria Letizia Jaccheri (NTNU, Norway)
Qualifying lecture: Open Source - State of the Art and Challenges Ahead.
Presentation: PDF
PhD Thesis
Research project with the title:
"Domain Model-Centric Distributed Development: An approach to semantics-based change impact management"
Supervisor: Prof. Arne Sølvberg
Abstract
Today's information systems engineering involves large number of stakeholders, wide geographical distribution and wide range of tools. Success in system engineering depends on effective human communication. Early understanding and modelling of the problem domain is a key to manage large scale systems and projects. This requires stakeholders to reach a certain level of shared interpretation of the domain referred throughout the development.
We propose a method for semantics driven change impact assessment. In our method, first a collaborative problem analysis is conducted. The problem analysis results in an agreed and committed common understanding of the problem domain, expressed in a conceptual domain model. The constructed conceptual domain-specific model is then actively used as a communication medium, e.g., to abstract development objects from representation format in order to explicate their semantics. Stakeholders browse the domain model and interactively associate to product fragments by selecting concept clusters that best describe the contents (intended meaning) of the product fragments.
Associations of the development objects with concepts from domain model, as well as the domain model itself constitute the basis for change impact assessment throughout the development. Every revision of a development object invokes change impact notifications that are either confirmed or rejected. Accumulated statistics are used to refine associations via the domain model to the direct dependency links among development objects.
The method has been implemented in a prototype system and has been evaluated in an experiment, where a set of test users has been provided with a problem domain description including a domain model and a set of development objects. The experiment was based on two real world cases. Users were asked to perform tasks using the prototype and two comparative tools. The method and prototype have been evaluated with respect to actual performance and users perceptions. The result shows actual effectiveness, perceived ease of use and usefulness comparing to other tools used in the experiment, as well as intention of the subjects to use the method in future.
Evaluation by the committee
Assessment of the qualifying lecture
"The lecture was presented in a very good manner. The topic was well researched and the presentation was structured and delivered in a competent style that engaged the audience. In summary the presentation was in conformance with the requirements as stipulated in the regulations."
Assessment of the dissertation
"The thesis addresses an important problem, i.e. that of distributed and collaborative software development and how to coordinate and manage development work in such a context. The thesis suggests means for supporting management and change impact prediction in connection with the various artifacts produced while developing an information system. The research problem is clearly stated and motivated. The research approach is one of suggesting and developing a solution which then shown to work in two (lab) experiments (proof of concept). The results are further evaluated by means of questionnaire investigating the perceived efficacy and usefulness of the tool, and hence whether the users would be inclined to use the tool, had it been commercially available.
The thesis accounts for state of the art in distributed and cooperative IS development, its methodology, tools and repositories, and then digs into how model and product fragments, their versions and configurations can be managed. It suggests a framework and method based on semantic relatedness to be arrived at by means of associating product fragments to a common domain model. The work finishes by the above mentioned evaluation. Six papers published during the years 2002 to 2005 constitute a peer reviewed presentation of the most important results.
The thesis demonstrates a good research framework within which the different components of the work have been carried out diligently and innovatively.
The contribution of the thesis is in the linking of domain ontology to emerging development artifacts by different development teams in a distributed environment. The thesis also makes a contribution to the methodological aspects for managing the process. A prototype tool is developed that implements the framework."
Evaluation Form: PDF