Kakia Chatsiou, University of Essex: Speaker at the TFNetwork Lightning Talk competition

Let's Get Physical - Autumn 2020 Virtual Conference
PGR SPOTLIGHT DAY

Held on 12th - 16th October 2020: Five days of Physics goodness on 

- Optics
- Wireless
- PGR Spotlight day
- Quantum 
- Data Science & AI

Find out more.

I am currently working as a Senior Researcher, at the ESRC Business and Local Government Data Research Centre. 

My research focuses on automated, quantitative methods of processing large amounts of textual and other forms of unstructured data – mainly political texts and social media – and the methodology of text mining for social science. I have published on applications of measurement and the analysis of text as data on machine learning methods and deep learning. I am also applying machine learning and natural language processing techniques to the analysis of public policy. My substantive research interests centre on resilience and the role of public policies and institutions at different levels of governance in shaping it. 

I am a member of the Natural Language and Information Processing Research Group, the Berkeley Initiative for Transparency in the Social Sciences.
LinkedIn

Recent publications:

• Chatsiou, K. and Mikhaylov, SJ., (2020). Deep Learning for Political Science. In: The SAGE Handbook of Research Methods in Political Science and International Relations. Editors: Curini, L. and Franzese, R., . SAGE Publications Ltd. 1053- 1078. 9781526459930
• Chamberlain, J., Turpin, B., Maged, A., Chatsiou, A. and O'Callaghan, K., (2020) Designing for Collective Intelligence and Community Resilience on Social Networks. Human Computation 
• Yannitell Reinhardt, G. and Chatsiou, K., (2019). Using community education interventions to build resilience and avert crises: how accidental dwelling fires decreased in Essex County, UK. Local Government Studies. 45 (3), 394-412

Political text classification using Neural Networks

We build a sentence-level political discourse classifier using existing human expert annotated corpora of political manifestos from the Manifestos Project (Volkens et al.,2020a) and applying them to a corpus of COVID-19 Press Briefings (Chatsiou,2020). We use manually annotated political manifestos as training data to train a local topic Convolutional Neural Network (CNN) classifier; then apply it to the COVID-19 Press Briefings Corpus to automatically classify sentences in the test corpus. We report on a series of experiments with CNN trained on top of pre-trained embeddings for sentence-level classification tasks. We show that CNN combined with transformers like BERT outperforms CNN combined with other embeddings (Word2Vec, Glove, ELMo) and that it is possible to use a pre-trained classifier to conduct automatic classification on different political texts without additional training.

Download slides (.pdf)