We build models that help machines understand and generate natural language - not only as text and speech, but in context. Combining knowledge graphs, neural networks and large language models, we create language resources and hybrid AI for research and society.
Discover our workUnderstanding and generating natural language
The “Artificial Intelligence and Language Technology” department focuses on creating models for understanding and generation of natural language by machine, using technologies of artificial intelligence such as knowledge representation in terms of knowledge graphs, and modelling and processing of language data using machine learning, including deep neural networks and large language models (based on Transformers).
Language in contextLanguage as a manifestation in context
We consider natural language not only as text and speech, but also as a manifestation in context. The context includes situations and factors from the outer world, such as images, world knowledge and knowledge about specific situations, knowledge of grammatical and semantic characteristics, as well as of the functions of the language itself.
Our resourcesPowering research and better AI
Language resources and technologies are mainly used for applications related to understanding and generation of natural language. They also support research in the Social Sciences and Humanities, as well as the construction of better large language models in generative AI. They are also used for detecting positive and negative stereotypes and biases in data, dealing with undesired errors in models, identifying missing types of knowledge, and training AI on small sets of examples.
Our approachHybrid: neural meets symbolic
In our work, we exploit hybrid approaches which combine neural network models with symbolic ones to achieve better synergy in both areas of research: artificial intelligence and natural language processing.
See the impactGood practices with social and scientific importance
Language resources and technologies are the basis for applications such as communication with artificial agents, descriptions of complex events, extraction and representation of knowledge from text, decision support, and augmentation of document content with additional knowledge. Through useful, accurate and fast interdisciplinary applications that benefit society, we create good practices with social and scientific importance.
Visit IICT