| _version_ | 1866901753508659200 |
|---|---|
| author | Sonar, Kartavya Satish |
| author_facet | Sonar, Kartavya Satish |
| contents | <p>This paper presents a systematic computational analysis of public discourse surrounding the United Kingdom's digital immigration infrastructure, focusing on the digital eVisa and the Biometric Residence Permit (BRP). We construct a corpus of 1,098 Reddit posts from seventeen communities and apply an NLP pipeline combining BERTopic-based topic modelling with a comparative evaluation of two Transformer-based emotion classifiers (GoEmotions and DistilRoBERTa). Ten coherent discourse themes are identified, with procedural friction accounting for over 55% of the corpus. We demonstrate that fine-grained emotion detection produces substantially richer policy-relevant insight than coarse classifiers, and introduce a fuzzy semantic matching layer linking discourse themes to UK legislative instruments. Results provide the first large-scale quantitative validation of concerns raised by civil society organisations about the UK's digital-by-default immigration programme.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_19651532 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Mapping Public Emotion in Digital Governance: A Comparative NLP Analysis of UK Immigration Discourse on Reddit Sonar, Kartavya Satish Natural language processing Natural Language Processing Sentiment Analysis topic modelling BERTopic digital governance UK immigration emotion detection computational social science <p>This paper presents a systematic computational analysis of public discourse surrounding the United Kingdom's digital immigration infrastructure, focusing on the digital eVisa and the Biometric Residence Permit (BRP). We construct a corpus of 1,098 Reddit posts from seventeen communities and apply an NLP pipeline combining BERTopic-based topic modelling with a comparative evaluation of two Transformer-based emotion classifiers (GoEmotions and DistilRoBERTa). Ten coherent discourse themes are identified, with procedural friction accounting for over 55% of the corpus. We demonstrate that fine-grained emotion detection produces substantially richer policy-relevant insight than coarse classifiers, and introduce a fuzzy semantic matching layer linking discourse themes to UK legislative instruments. Results provide the first large-scale quantitative validation of concerns raised by civil society organisations about the UK's digital-by-default immigration programme.</p> |
| title | Mapping Public Emotion in Digital Governance: A Comparative NLP Analysis of UK Immigration Discourse on Reddit |
| topic | Natural language processing Natural Language Processing Sentiment Analysis topic modelling BERTopic digital governance UK immigration emotion detection computational social science |
| url | https://doi.org/10.5281/zenodo.19651532 |