JPC1

Within the course, the student will become familiar with the principles, options and titles for research, development and innovation. They will also acquire the ability to work as a team member and perform partial tasks.

2025 Goals --- Metaphor & Metonymy

We want to document lexical metaphors in multiple languages.

Read:

• ChainNet: Structured Metaphor and Metonymy in WordNet (Maudslay et al., LREC-COLING 2024)
• Metonymy is more multilingual than metaphor: Analysing tropes using ChainNet and the Open Multilingual Wordnet (Bond and Maudslay, GWC 2025)

Who is doing what

• LCM and patterns
  • Barbora Mahdalová - bmahdalova - Look at xlingual morphological patterns
  • Dominik Fránek - dfranek161 - Look at xlingual morphological patterns
• Different POS
  • Jakub Pokorný - Extend to different POS
  • Nikola Zdařilová --- --- Extend to different POS
• Contrast to Unimet
  • Aneta Zatloukalová- anetazatloukalovaa --- Compare and contrast with UniMet
  • Bára Havlíčková- barahavlis - Compare and contrast with UniMet
• Lexical Conceptual Mapping
  • Vojtěch Mahdal - Vmahdal - lexical conceptual mapping
  • Lea Karolína KOMÁRKOVÁ.
• Wordnet
  • Francis BOnd - fcbond - investigate the behaviour of metaphors at scale.

Tasks

Wordnet integration

• import chainnet to (English) wordnet
• project to other languages
• visualize
• compare to corpus
• look at sentiment
• Look for errors
  • find exceptions to patterns
  • evaluate and fix
  • mainly direction, but can be type

Lexical Conceptual Mapping

• map metaphor pairs to hypernyms
• look at patterns
• read MetaNet (or look for more up-to-date source)
• http://araw.mede.uic.edu/~alansz/metaphor/METAPHORLIST.pdf
• try to map
  • try to generate labels for new ones

Compare and contrast with UniMet

• see where we differ and why
• evaluate
• write script to merge to wordnets (like task A)
• https://github.com/kbatsuren/UniMet

Extend to different POS

• project from nouns using derivational links
  • hand annotate a subset to see what we are missing
  • look at other approaches
• see what her have in other data sets (like MetaNet)

Look at xlingual morphological patterns

• often a zero derivation in one language will be marke din another

• use this to type the links

• use wordnet

• use wiktionary ** find hypernym from definition ** look at senses with different hypernyms (use PERSON/ANIMAL but make it programmable)

• focus on a language or focus on multilingual?

• look at positive negative?

• look at borrowing?

• maybe also look at morpholical links

• also at lexical-conceptual mapping

Parse DBnary

• try to find hyponyms
• try to match trope classes

Extend wordnet with DBnary

• https://kaiko.getalp.org/about-dbnary/
• read data (with RDF lib)
• attempt to map to wordnet synsets
• use definition, use translations