The “jump and stay” method is described here: The “jump and stay” method to discover proper verb centered constructions in corpus lattices
The foundation, the double cube model is described here: A lattice based algebraic model for verb centered constructions
Please type:
make all
to run the algorithm (described in section 5)
- using verb hagy (allow) on train data
to get the full output which Fig. 4 is based on
(in file named
hagy.train.out3); and - using verbs húz (draw/pull) and vet (cast/throw) on test data
to get the results presented in Table 1. in the [paper](RANLP 2019...)
(in files
huz.test.out3.pVCCandvet.test.out3.pVCC).
Tested on Debian Linux. (May work on other operation systems...)
Requirements: python 3 make uni2ascii (if not available: sudo apt-get install uni2ascii)
This method is considered language independent. See: https://github.com/sassbalint/double-cube-jump-and-stay-multilingual
If you want to use this, please cite the above papers and contact me. :) No warranty, sorry.