pmb2tsv

pmb2tsv is a collection of scripts to convert data from the Parallel Meaning Bank (PMB) into column-based files.

Input Data

Please download the PMB 3.0.0 and extract the directory pmb-3.0.0 into a data directory in the root directory of this repository (or symlink it).

Software Dependencies

Scripts to convert the files are mostly found in this repository; however, the following software needs to be present on the system:

• Python 3 – the python3 executable should be on your $PATH.
• Produce – the produce executable should be on your $PATH.
• SWI-Prolog 7 or higher – the swipl executable should be on your $PATH.
• GNU Parallel – the parallel executable should be on your $PATH.

Conversion

Now use the produce command to convert the desired portions of the PMB to TSV files. For example, to get all training data, run:

produce out/pmb-3.0.0-{en,de}-{bronze,silver,gold}-train.tsv
produce out/pmb-3.0.0-{it,nl}-{bronze,silver}-train.tsv

And to get all gold development data, run:

produce out/pmb-3.0.0-{en,de,it,nl}-gold-dev.tsv

This example will generate 8 TSV files, one per language and part. They contain the converted sentences, separated by empty lines, one token per line with the following tab-separated columns:

1. Token number within sentence
2. Token form
3. Lemma
4. Semantic tag
5. WordNet sense
6. CCG dependency head
7. DRS fragment

Warning: for a small number of CCG derivations, especially some that are not fully corrected, dependency and role extraction will fail. The corresponding columns will be empty/missing. In extremely rare cases a dependency non-tree (a cyclic graph) may be extracted.

For details on the conversion from CCG derivations to dependency trees, see

Kilian Evang (2020): Configurable Dependency Tree Extraction from CCG
Derivations. Proceedings of the Universal Dependencies Workshop.

To reproduce the experiments from that paper, checkout out the evang-2020-configurable tag and run:

produce pmb-3.0.0-{en,de,it,nl}-gold-{p00,p01}.eval