• FAIR (Findability, Accessibility, Interoperability and Reusability) metadata: Solutions to add a semantic layer to the data management framework, which is linked to the metadata of the underlying data. Their main purpose is to harmonize different data and metadata schemas as well as different vocabularies and ontologies. Thus, both data and metadata gain explicitly semantic meaning.
• Data in context: Methods and tools that repeatedly reuse a common knowledge base, contexts for data enrichment and thus improve data quality. For example, enriching a company’s data with a context from external sources such as news and social media deliver additional information about factors that influence its success. Later on, the exploration of data in context allows for the identification of external factors, that can be used to guide future decisions.
• Data fabric: Methodologies and tools for development “fabric” for data integration and linking. Data fabrics produce federated data and allow for the translation of the existing data models into semantic knowledge models such as ontologies and taxonomies. They could be implemented using knowledge graphs in which all rules for the meaningful and dynamic linking of business objects are stored. Thus, the advantages of Data Lakes and Data Warehouses are complemented with the advanced linking methods of Semantic Graph Technologies.
• Explicit semantics: Methods and tools that make the semantics of data explicit, which in turn becomes accessible, machine-readable and portable.
• Store and manage data as knowledge: Exploration and application of knowledge graphs as a powerful knowledge management system.
• Data quality: Methods and tools that validate data consistency and provide repair mechanisms on it as well as perform quality assurance or sanity check after enrichment, so that the quality and completeness can be assessed.

• Center for the Study of Democracy – disinformation and policy research;
• Ontotext – semantically-enriched database of debunks, AI tools for content verification;
• Prof. Nicoleta Corbu (College of Communication and Public Relations, SNSPA) – extensive research in the field of political communication, with a focus on media effects, disinformation, and news consumption.
• Funky Citizens – media literacy in the Balkans, running the factual.ro political fact-checking platform in Romania;
• Dr Ralitsa Kovacheva – journalism research in hybrid threats;
• Dr Bissera Zankova – Media 21 Foundation – media freedom; media literacy; law.

Under the leadership of Prof. Kalina Bontcheva, we established a new regional multi-disciplinary alliance of researchers, fact-checkers and other experts, to act as a catalyst and an amplifier of national and regional actions against disinformation.

Key staff

Assoc. prof. Milena Dobreva is the Research Leader on disinformation at GATE Institute.

Prof. Kalina Bontcheva is the head of the Natural Language Processing (NLP) group at the University of Sheffield. She has coordinated the FP7 PHEME project on rumour and disinformation analysis and is the scientific director of the H2020 WeVerify project on collaborative content verification. Since 2014, Prof. Bontcheva has been researching methods for disinformation analysis combining NLP methods for analysing the textual content with machine learning models for credibility, veracity, and stance detection. These are combined also with the analysis of the mention and hashtag-based networks, in the context of misinformation spread, voting intentions and online abuse.

Prof. Ivan Koychev is the Data Analytics research group leader and leads the MSc. Program in Information Retrieval and Knowledge Discovery. He gained experience in a number of academic and industrial projects including contribution to seven EU funded projects. His main research interests are in the areas of Machine Learning, AI, Information Retrieval, Data/Text Mining and their applications to disinformation analysis. He also brings in a strong track record in Natural Language Processing in Bulgarian and English. His most recent publications have focused on deep learning methods for check worthiness analysis (i.e. whether a given post needs to be fact-checked or not), detection of deception in political debates, verification of claims about images, and detection of toxicity in news articles.

Selected publications

1. Carolina Scarton, Diego F. Silva, Kalina Bontcheva. Measuring What Counts: The case of Rumour Stance Classification. Proc. of the 1st Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 10th International Joint Conference on Natural Language Processing (AACL-IJNLP), 2020, http://aacl2020.org/program/accepted/
2. Alex Nikolov, Giovanni Da San Martino, Ivan Koychev, Preslav Nakov, Team Alex at CLEF CheckThat! 2020: Identifying Check-Worthy Tweets With Transformer Models, CLEF 2020, 22-25 September 2020, Thessaloniki, Greece., https://arxiv.org/abs/2009.02931
3. Dimitrina Zlatkova, Preslav Nakov, Ivan Koychev, Fact-Checking Meets Fauxtography: Verifying Claims About Images, Proc. EMNLP-IJCNLP, Publisher: Association for Computational Linguistics, 2019, doi:10.18653/v1/D19-1216, https://arxiv.org/abs/1908.11722
4. Michal Lukasik, Kalina Bontcheva, Trevor Cohn, Arkaitz Zubiaga, Maria Liakata, Rob Procter. Gaussian processes for rumour stance classification in social media. ACM Transactions on Information Systems (TOIS) 37 (2), 1-24. 2019. https://doi.org/10.1145/3295823
5. Ahmet Aker, A Sliwa, F Dalvi, Kalina Bontcheva. Rumour verification through recurring information and an inner-attention mechanism. Online Social Networks and Media 13, 100045. 2019. https://doi.org/10.1016/j.osnem.2019.07.001
6. Todor Mihaylov, Tsvetomila Mihaylova, Preslav Nakov, Lluís Màrquez, Georgi D. Georgiev, Ivan Koychev, The dark side of news community forums: opinion manipulation trolls, Internet Research, vol:28, issue:5, 2018, pages:1292-1312, ISSN (online):1066-2243, https://doi.org/10.1108/IntR-03-2017-0118, https://www.emerald.com/insight/content/doi/10.1108/IntR-03-2017-0118/full/html

Contact

Milena Dobreva, milena.dobreva@gate-ai.eu

Data Analytics is applied using online and scalable machine learning algorithms, which are able to continuously update the learned models and to work on distributed systems. At the same time, Real-time Data Analytics dramatically changes the ways systems can use data to predict outcomes and suggest alternatives. Instead of putting together conjectures based on a series of past events and recent scans, systems working in real-time can deliver insights on what is going on right now. In addition, new machine-learning systems have the ability to explain their rationale, characterize their strengths and weaknesses, and convey an understanding of how they will behave in the future.

Current Research Directions:

• Machine Learning for Big Data
• Natural Language Processing for low resource languages and concrete application domains
• Explainable AI

Targeted applications:

• e-health
• web data analytics
• disinformation detection etc.