Venelin Kovatchev

Venelin Kovatchev

Postdoctoral Researcher
University of Birmingham

My primary research areas are natural language processing and computational semantics. I am interested in the meaning of complex linguistic expressions, (common sense) reasoning, interpretability, interdisciplinarity, and a scientifically sound methodology.

I currently use state-of-the-art NLP for the automatic scoring of psycho-linguistic theory of mind tests. I also design resources and systems for Natural Language Inference in Spanish and Catalan.

I got my PhD (2020) and my M.Sc. (2015) at the University of Barcelona. Prior to that, I worked as a software developer.

Automatic Scoring of Mindreading

Mindreading, or theory of mind, is the ability to understand the feelings, beliefs, and desires of others. Individual differences in children’s mindreading are linked with both social and academic outcomes and children’s wellbeing. Furthermore, difficulties with mindreading are linked with a range of mental health problems and neurodevelopmental conditions.

As a postdoctoral researcher at the University of Birmingham, I am currently working on an interdisciplinary project involving computer science, linguistics, and developmental psychology. I have developed an end-to-end system for the automatic scoring of the open-ended responses to standardized theory of mind tests.

first service
second service

Decomposition of Textual Meaning Relations

The automatic processing of textual meaning relations such as textual inference (entailment), contradiction, paraphrasing, and semantic similarity are part of the larger framework of natural language understanding (NLU) and are often used as an evaluation benchmark for novel machine learning architectures.

My research on this topic improves our understanding of the working of textual meaning relations, the multiple linguistic and reasoning phenomena involved in them, and the way automatic systems can process them. I also identify important limitations in the existing resources and systems.

Dataset Creation and In-Depth System Evaluation

The creation of large-scale high-quality datasets and the in-depth evaluation of ML systems via statistical measures and human experts is essential for the successful application of automated soltions.

In my research research I have explored and compared different strategies for creation and annotation of corpora: expert corpus creation, crowd-sourcing, and data augmentation. I also propose methodologies for detailed evaluation and semi-automatic error analysis of automated systems.

first service

List of Publications

Kovatchev, V., Smith, P., Lee, M, and Devine, R., “Can Vectors Read Minds Better Than Experts? Comparing Data Augmentation Strategies for the Automated Scoring of Children’s Mindreading Ability” at Proceedings of the Joint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (ACL-IJCNLP), 2021

Kovatchev, V., Smith, P., Lee, M, Grumley Traynor, I., Luque Aguilera, I., and Devine, R., “What is on your mind?” Automated Scoring of Mindreading in Childhood and Early Adolescence at Proceedings of the 28th International Conference on Computational Linguistics (COLING), 2020

Hossain, M. M., Kovatchev, V., Dutta, P., Kao, T., Wei, E., Blanco, E., "An Analysis of Natural Language Inference Benchmarks through the Lens of Negation" at Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020

Kovatchev, V., Gold, D., Martí, M. A., Salamó, M., and Zesch, T., "Decomposing and Comparing Meaning Relations: Paraphrasing, Textual Entailment, Contradiction, and Specificity" at Proceedings of The 12th Language Resources and Evaluation Conference, 2020

Ikauniece, I., Kovatchev, V., Puertas, E., "Spanish and Catalan Polarity Detection in Student Satisfaction Surveys", at Proceedings of the 21st International Conference of the Catalan Association for Artificial Intelligence, 2019 Best Poster Award

Kovatchev, V., Martí, M. A., Salamó, M., and Beltran, J. "Qualitative Evaluation Framework for Paraphrase Identification", at Proceedings of the 12th Recent Advances in Naural Language Processing conference, 2019

Gold, D., Kovatchev, V., and Zesch, T. "Annotating and analyzing the interactions between meaning relations", at Proceedings of the 13th Language Annotation Workshop, 2019

Martí, M. A., Taulé, M, Kovatchev, V., and Salamó, M. "DISCOver: DIStributional approach based on syntactic dependencies for discovering COnstructions", at Corpus Linguistics and Linguistic Theory, 2019

Kovatchev, V., Martí, M. A., and Salamó, M., "WARP-Text: A Web-Based Tool for Annotating Relationships Between Pairs of Texts", at Proceedings of the 27th International Conference on Computational Linguistics, 2018, System Demonstrations

Kovatchev, V., Martí, M. A., and Salamó, M., "ETPC - A Paraphrase Identification Corpus Annotated with Extended Paraphrase Typology and Negation", at Proceedings of the 11th edition of the Language Resources and Evaluation Conference, 2018

Kovatchev, V., Salamó, M., and Martí, M. A., "Comparing models of distributional semantics for identifying groups of semantically related words", at Procesamiento de Lenguage Natural, 57, 2016

Teaching

Programming for Data Science (2020-2021)
co-Lecturer
Master Program of Data Science,
School of Computer Science,
University of Birmingham

Introduction to Python programming (2018-2019; 2019-2020; 2021-2022)
Lecturer
Language Technology Service (STEL),
Facilty of Filology and Communication,
University of Barcelona

Machine Learning for NLP (2018, 2019, 2021)
Senior Monitor
Lisbon Machine Learning School (LxMLS)

Introduction to NLP (2016-2017; 2017-2018)
co-Lecturer
Master Program of Data Science,
Faculty of Mathematics and Computer Science,
University of Barcelona

Contact Information

Email: v.o.kovatchev [locative] bham [dot] ac [dot] uk
Twitter: @sintelion
LinkedIn: Venelin Kovatchev
GitHub: venelink

Curriculum vitae

You can find my most recent academic CV here