Tracing requirement objects as an information retrieval task

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Master's thesis

Date

2020-08-18

Department

Major/Subject

Autonomous Systems

Mcode

ELEC3055

Degree programme

Master's Programme in ICT Innovation

Language

en

Pages

66+8

Series

Abstract

In large requirement databases, tracing of different objects to each other, e.g. higher-level requirements to lower-level requirements, or requirements to their verification methods, can be a tedious job. With numerous objects in the database the selection of the corresponding object from lists can take a long time. In this thesis standard information retrieval (IR) methods, in particular multiple variants of vector space modelling, are applied in order to provide a shortlist of a few objects, which are predicted to be relevant, this way speeding up the selection process. The aim of the thesis is to demonstrate the usage of such IR system on a real-life example requirement data set, providing an end-to-end solution from processing the relevant data to showing the shortlist on a GUI view. The separation of the data to train and validation subsets and the setup of a relevant evaluation metric is also essential in order to benchmark future developments.

Description

Supervisor

Oulasvirta, Antti

Thesis advisor

Hujanen, Jaakko
Leiva, Luis

Keywords

information retrieval, vector space model, requirement tracing, automated requirement tracing

Other note

Citation