Publications

We investigate how effectively automatic named entity recognition can recover entities selected by investigators in real criminal cases. Two homicide cases and one synthetic case were processed with five NER models, and extracted entities were matched to investigator labels using a lightweight entity resolution and similarity scoring method. Police trained models aligned best with investigator annotations, while general-purpose models produced larger but less relevant entity sets. Manual validation confirmed that many labels are not text recoverable, establishing a realistic upper bound on NER performance. The results indicate that extraction and investigative relevance are distinct tasks, motivating a two-stage pipeline and future work on relevance modelling and improved resolution.

Case management systems (CMSs) are essential to modern criminal investigations but remain underexplored compared to systems for predictive policing and digital forensics. This study examines how investigators use CMSs, which capabilities remain insufficient, and how these systems can better support their work. Through interviews with practitioners across European jurisdictions, we identify an investigative workflow of six concurrent dimensions and propose a two-tier model distinguishing preparation and exploration for knowledge discoveries. While CMSs effectively manage actions and decision-making, critical gaps persist for knowledge discovery, such as text analysis, hypothesis testing, and information retrieval. We derive functional requirements by workflow dimensions, offering guidance on which capabilities should be embedded in core systems and which are better delivered as modular services. These findings inform the design of CMSs that combine stable infrastructure with flexible tools.

This study investigates the prevalence and patterns of irregular deposit practices and legal breaches concerning tenancy deposits in Norway’s private rental housing sector. Irregularities in handling tenancy deposits can adversely affect tenants, particularly vulnerable groups like students and low-income families. By analysing transaction data from DNB, Norway’s largest bank, covering January 1, 2014, to December 31, 2023, the research reveals that approximately 24.5–42.6 % of rental units experience irregular deposit practices. Many deposits are not stored in legally required accounts, and about 85 % of deposit repayments were round figures, indicating that interest accrued on deposits was often not paid back to tenants. These breaches of the Norwegian Tenancy Act highlight significant concerns within the private renting sector. As the first systematic investigation of irregular deposits in Norway’s rental market, this study provides a foundation for future research and interventions to protect tenant rights and improve compliance with deposit regulations. The findings emphasize the necessity for legislative reforms to strengthen tenant protections and ensure proper deposit-handling .

Purpose – The purpose of this paper is to develop, describe and validate a machine learning model for prioritising which financial transactions should be manually investigated for potential money laundering. The model is applied to a large data set from Norway’s largest bank, DNB. Design/methodology/approach – A supervised machine learning model is trained by using three types of historic data: “normal” legal transactions; those flagged as suspicious by the bank’s internal alert system; and potential money laundering cases reported to the authorities. The model is trained to predict the probability that a new transaction should be reported, using information such as background information about the sender/receiver, their earlier behaviour and their transaction history. Findings – The paper demonstrates that the common approach of not using non-reported alerts (i.e. transactions that are investigated but not reported) in the training of the model can lead to sub-optimal results. The same applies to the use of normal (un-investigated) transactions. Our developed method outperforms the bank’s current approach in terms of a fair measure of performance. Originality/value – This research study is one of very few published anti-money laundering (AML) models for suspicious transactions that have been applied to a realistically sized data set. The paper also presents a new performance measure specifically tailored to compare the proposed method to the bank’s existing AML system.

This bachelor thesis examines where the upper boundary of surveillance by Norwegian banks should lie in efforts to combat money laundering and terrorist financing. It considers regulatory requirements, technological change, privacy and the resources devoted to monitoring, drawing on criminological literature and interviews with relevant decision-makers and supervisory actors.

The Arctic is increasingly becoming an important area in Industry, Science, Tourism, and Education. With the increased activity, the number of accidents are increasing and causing a demand for better safety management. It is found that little research into the topic is previously conducted. In this Thesis, three datasets with incidents in the Svalbard region have been investigated, using a combination of accident concentration analysis, cause analysis and preventive strategies analysis. It is found that most accidents either occur during the spring or late summer period at Svalbard due to the scooter and maritime season respectively. Five incident types that are characteristic for the Arctic region has been identified; Assistance person, fall due to ice or snow, scooter incidents, unintentional discharge of weapons, and events that can not be foreseen. It is found that the operations at Store Norske Spitsbergen Kullkompani do not show any significant differences caused by being in the Arctic. The safety operations at the University of Svalbard are shown to have multiple challenges in both safety management systems and incident concentrations, and it is a combination of chance and good emergency response at UNIS that no fatalities have occurred. The UNIS operations are also shown to be adapt to ensure safe field activity despise of this.

During the spring of 2014 fieldwork in three different locations on Svalbard was conducted in order to investigate the application of electromagnetic methods on sea ice and ice ridges. Two validation grids, one transect across Sveasundet, a walkabout, an ice ridge and 8 ice cores have been measured. Results show an average thickness less than one meter for all but the ice ridge, which was 1,9 meters at the thickest point. Simulation based calculation of the relationship between the measured apparent conductivity and the level ice thickness, using MarcoAir, show great accuracy and potential. The sea ice conductivity profile is shown to have great effect on the calculated thickness, and this becomes more and more apparent as the thickness of the ice increases. This effect is shown, by using a simple quasi-2D model of an ice ridge, to hold for ice ridges. Investigation into the potential methods of estimating internal structures of ice ridges show that there are great potential in using simulation and modelling in combination with inversion, in order to assess these properties.