Danish Asylum Adjudication using Deep Neural Networks and Natural Language Processing

Satya Mahesh Muddamsetty*, Mohammad Naser Sabet Jahromi, Thomas B. Moeslund, Thomas Gammeltoft-Hansen

*Kontaktforfatter

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

198 Downloads (Pure)

Abstract

The Danish asylum adjudication procedure is a two-tiered
system, with the Immigration Service making initial determinations and the Danish Refugee Appeals Board (RAB) automatically appealing cases that are rejected. This study aims to employ a deep neural network(DNN)- based Natural Language Processing (NLP) pipeline to predict asylum decision-making outcomes using a dataset of over 15,515 Danish asylum decisions provided by the Danish Refugee Appeals Board (RAB) between January 1995 and January 2021. This research seeks to improve
the performance and effectiveness of decision-making in asylum cases by addressing key challenges, such as modeling the asylum decision-making problem using NLP-based DNNs and dealing with class imbalance issues.
Our preliminary results indicate that DNN-based NLP predictive models are capable of learning meaningful representations of asylum cases with high precision and recall, particularly when class weights are considered than the baseline DNN model.
OriginalsprogEngelsk
TitelProceedings of the Seventeenth International Workshop on Juris-Informatics 2023 (JURISIN 2023)
Publikationsdatojun. 2023
Sider92-105
StatusUdgivet - jun. 2023
Begivenhed17. International Workshop on Juris-informatics
(JURISIN 2023)
- Kumamoto, Japan
Varighed: 5 jun. 20236 jun. 2023
https://research.nii.ac.jp/~ksatoh/jurisin2023/

Konference

Konference17. International Workshop on Juris-informatics
(JURISIN 2023)
Land/OmrådeJapan
ByKumamoto
Periode05/06/202306/06/2023
Internetadresse

Fingeraftryk

Dyk ned i forskningsemnerne om 'Danish Asylum Adjudication using Deep Neural Networks and Natural Language Processing'. Sammen danner de et unikt fingeraftryk.

Citationsformater