This repository contains the datasets POEMS and TSV described in the paper Metaphor detection for German Poetry (2019). TSV is based on the English dataset of Tsvetkov et al. (2014).
The folders contain the following data:
- poems: POEMS dataset (train and test)
- tsv-translated: TSV dataset (train and test)
In both folders, the files are organized as follows:
-
Training data: files starting with
train_ -
Test data: files starting with
test_ -
Metaphorical instances: files ending with
_met -
Non-metaphorical instances: files ending with
_nonmet -
Ambiguous instances: files ending with
_ambig
Each file contains one lemmatized adjective-noun pair per line.
Please note that ambiguous instances were not used in experiments.