Meticos-Project / Project Blog  / Towards Understanding of User Perceptions for Smart Border Control Technologies using a Fine-Tuned Transformer Approach

Towards Understanding of User Perceptions for Smart Border Control Technologies using a Fine-Tuned Transformer Approach

Introduction and Background

SBC technologies are fully or semi-automated technologies with the purpose of verifying traveller’s identity without any human (i.e., border guard) intervention [1]. Despite the potential benefits of the SBC technologies at BCPs, users (i.e. travellers) perceptions of acceptability of the technologies may change with actual experience of users with them.

We have a focus on finding out the reasons for:

  • why acceptance of SBC technologies depend on user perceptions,
  • changes in user perceptions for SBC technologies, and
  • changes in user perceptions after users have actual experience with SBC technologies

through questionnaires and social media data analysis

Table 1. State-of-the-art studies for extracting user perceptions from social media data

Authors Social Media Platform Data Perceptions Technique Type Results
Khan et. al. [2] Yelp Yelp Challenge Dataset (2835065 user reviews) Rating perceptions to restaurants, ecommerce Users Rating Classification using Machine Learning (NaiveBayes, RandomForest, DecisionTree, SVM, KNN, MLP) Supervised 96% NaiveBayes
Xuan et. al. [3] Facebook Facebook Posts (37426 in health, technology, travel, finance, sports) Users interests and trends (i.e. health, tech, travel, sports) Users Trends Classification using Machine Learning (NaiveBayes, DecisionTree, SVM), Deep Learning (CNN) Supervised 91% CNN
Alshamrani et. al. [4] Twitter Tweets (26 million b/w 01/10/2019 and 30/04/2020) users behaviours and trends before and after COVID-19 Deep Learning for Sentiment Analysis (DNN, CNN, LSTM, BERT), LDA for Topic Modelling Supervised 87% BERT
Rosario et. al. [5] Twitter and surveys N/A users experiences and perceptions for e- learning tools Deep Learning (CNN, BiLSTM, Transformers), Google, Microsoft APIs Supervised 61.7% Transformer
Molinari et. al. [6] Twitter Tweets (13800 related to URBAN- GEO BIG DATA sensing) Investigate citizens perceptions of mobility services for urban geo sensing Machine Learning (NaiveBayes) Supervised N/A


Methods and Materials












This study analyzed user tweets in order to understand user perceptions for SBC technologies by combining BERT model with the Flair labelling technique.

Advantages of the proposed system

– automatic sentiment labels generation when there is no labelled data exists

– applying the proposed study for extracting sentiments to understand user perceptions for SBC technologies

The highest performing model was the combination of Flair and BERT with 79% accuracy on the IMDB test dataset

The proposed approach can be customized for use in other domains where raw data exists, however, there is no processed and labelled data available.



  1. RDU Best practice operational guidelines for automated border control (abc) systems.European Agency for the Management of Operational Cooperation, Research and Development Unit,. https://bit. ly/2KYBXhz Ac-cessed, 9(05):2013, 2012.
  2. ShahidaKhan, Kamlesh Chopra, andPratyush Brand review prediction using user sentiments: Machine learning algorithm. In2nd International Conference on Data, Engineering and Applications (IDEA),pages 1–8. IEEE, 2020.
  3. Xuan Truong Dinh and Hai Van A proposal of deep learning model for classifying user interests on social networks. InProceedings of the 4th International Conference on Machine Learning and Soft Computing, pages10–14, 2020.
  4. Sultan Alshamrani, Ahmed Abusnaina, Mo-hammed Abuhamad, Anho Lee, DaeHunNyang, and David An analysis of users engagement on Twitter during the covid- 19 pandemic: Topical trends and sentiments. In International Conference on Computational Data and Social Networks, pages 73–86. Springer, 2020.
  5. Rosario Sanchis-Font, Maria Jose Castro-Bleda, Jose-Angel Gonzalez, Ferran Pla, andLlu ́ıs-F Cross-domain polarity mod-els to evaluate user experience in e-learning. Neural processing letters, pages 1–17, 2020.
  6. ME Molinari, D Oxoli, CE Kilsedar, andMA User geolocated content analysis for urban studies: investigating mobility perception and hubs using Twitter. InISPRSTechnical Commission IV Symposium 2018,volume 42, pages 439–442, 2018

Sarang Shaikh1, Sule Yildirim Yayilgan1, Erjon Zoto1, Mohamed Abomhara1 1Dept. Of Information Security and Communication Technology (IIK), Norwegian University of Science and Technology (NTNU), Gjøvik, Norway