Skip to content

icsi-berkeley/framenet-multilingual-alignment

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

60 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Multilingual Framenet

If you want to experiment with ViToXF, our alignment visualization tool, you can download the most recent version under this repo Releases (link). Please refer to the user's guide and if you have any question, feel free to open an issue. Thank you!

We're still working on this project, documentation is still not completed and we plan to add more languages and expand the techniques of the existing ones.

Preparing FrameNet data

  • Create a folder named "data" in the project's root
  • Inside "data", you should add a folder for each database you want to align. Some names, such as bfn, chineseefn and swedishfn, are already associate to specific data loaders, i.e., they are expected to be in the formats provided by those responsible for those databases.
  • In any case, each database folder in data can be associated with a specific loader. To do that, navigate to alignment/fnalign/loaders.py and include your database folder name in the supported_dbs function.
  • By default, you should consider using the FNLoader, as it supports the same format as the Berkeley FN 1.7 data release.

Preparing MUSE data

  • To run alignment methods that use MUSE data, you need to also download the embeddings from here
  • Create and the data/muse if it doesn't exists and move all the relevant .vec files there.

Running the alignment

  • Make sure you have Conda installed (https://docs.conda.io/en/latest/)
  • Run the following commands on the project's root:
  • conda env create -f environment.yml
  • conda activate mlfn
  • Before running the alignemnt, open alignment/main.py and comment every line related to scoring techniques that you don't want to run (In a future version, this will be done using a CLI)
  • Finally, run:
  • python3 ./alignment/main.py

After running the output files will be in the out folder on the project's root.

Running the visualizer