Skip to main content

Master's lecture in Computer Science - Hjalti Daníelsson

When 
Wed, 18/12/2019 - 14:00 to 16:00
Where 

Veröld - Hús Vigdísar

Room 103

Further information 
Free admission

Master's student: Hjalti Daníelsson

Title: Software for automatic term extraction on domain-specific Icelandic texts.
___________________________________________
Faculty: Faculty of Industrial Engineering, Mechanical Engineering and Computer Science

Advisor: Ebba Þóra Hvannberg, Professor at the Faculty of Industrial Engineering, Mechanical Engineering and Computer Science

Other member of the masters committee: Helgi Þorbergsson, Associate Professor at the Faculty of Electrical and Computer Engineering

Examiner: Hannes Högni Vilhjálmsson, Associate Professor at Reykjavík University

Abstract
This thesis presents software that performs automatic term extraction on domain-specific Icelandic-language texts. The conditions that accompany software development in this kind of context are explored, including current theories on extraction methodologies, the most impactful features of the Icelandic language, and the availability of existing software in this field. In addition, we describe the part our software plays in a terminology project launched by The Árni Magnússon Institute for Icelandic Studies (AMI), and the conditions to be considered in that context. In light of the above, we describe why and how we decided to
construct the aforementioned software from scratch. The software's design is subsequently delineated. We review its primary sections - preprocessing, linguistic processing, statistical processing, and output - and describe each one in terms of what methods we chose and why, and how they were implemented. Testing methodologies are discussed, after which we review the results of testing the software's functionality. We find that the software performs well in the role for which it was intended. The methodology it employs delivers dependable results, and the support programs it relies on for certain aspects of its functionality are found to perform satisfactorily. Lastly, we review the development process, and discuss future possibilities for additional functionality in this kind of program.