Institute of Formal and Applied Linguistics
Charles University, Czech Republic
Faculty of Mathematics and Physics
Master or Bachelor Thesis at ÚFAL
If you think about a thesis concerning language technologies, then the best way to proceed is to contact us at thesis@ufal.mff.cuni.cz (feel free to write either in Czech, Slovak or English). Alternatively, if you prefer contacting individual prospective supervisors, you are welcome to do so.
- If you have a topic or area of interest of your own, we will be very glad to hear about it!
- We also list the overview of areas we at ÚFAL are actively working on, so please share with us what kind of language technologies do you find interesting.
Machine Translation
- Best Systems for Czech and Other Languages Translation
- Offline Machine Translation Bergamot
- Live Speech Translation ELITR Whisper-Streaming Details
Topics include new, promising and challenging applications, such as live post-editing by humans or LLMs, cross-lingual dialogues, multilingual summarization, quality estimation, applications of LLMs, etc.
Prospective supervisors: Dominik Macháček, Ondřej Bojar
- Multimodal Translation Multi30k
Dialogue Systems PapersWithCode
- Task-oriented Dialogue Systems PapersWithCode Details
Interactive text-based or speech-based systems that help fulfill tasks (e.g. booking tickets, searching for bus connections, restaurant recommendation). Works on this can include:
- End-to-end neural systems (see e.g. this shared task challenge or this dataset)
- Dialogue system components: Language understanding, dialogue management
- Open-domain Dialogue PapersWithCode Details
Chat about anything with Large Language Models
- Automatic Speech Recognition UWebASR PapersWithCode
- Text to Speech PapersWithCode
Language modeling PapersWithCode
Natural Language Generation PapersWithCode
- Czech poetry generation EduPo
- Data-to-text generation PapersWithCode NG-NLG Details
This can include response generation for dialogue systems as well as data-to-text generation (weather reports, sports reports etc.) or surface realization (generating text from semantic/syntactic trees).
Morphosyntactic Analysis
- Part of Speech Tagging, Lemmatization UDPipe MorphoDiTa UD PDT PapersWithCode Details
Current Czech best results on arXiv.org.
- Dependency Parsing UDPipe UD PDT PapersWithCode Details
Current Czech best results on arXiv.org.
- Derivational Morphology DeriNet
Natural Language Understanding
- Coreference Resolution CorefUD CRACSharedTask PapersWithCode
- Semantic Parsing PDT PapersWithCode
- Discourse Analysis PDT Details
Discourse analysis includes among others:
- Discourse Connectives
- Implicit Discourse Relations
- Question Answering SQuAD PapersWithCode
- Summarization SumeCzech PapersWithCode
Named Entities
- Named Entity Recognition NameTag CNEC PapersWithCode Details
Current Czech best results on arXiv.org.
- Named Entity Linking PapersWithCode
- Relationship Extraction PapersWithCode
NLP Applications
- Grammar Error Correction Korektor AKCES-GEC PapersWithCode
- Optical Character Recognition Details
Detailed description of the thesis Adaptive Handwritten Text Recognition
Music Processing
Many more music-related topics are possible. See what others are working on at the Prague Music Computing Group page.