##UPF-Cobalt Metric for Machine Translation Evaluation UPF-Cobalt is a MT evaluation metric that exploits alignment and syntactic context to assess MT quality.
###Installation
- Install Stanford Parser (http://nlp.stanford.edu/software/corenlp.shtml), version 3.4.1
- Download dependency-based word vectors from https://levyomer.wordpress.com/2014/04/25/dependency-based-word-embeddings/ (optional)
- Download upf-cobalt
git clone https://github.com/amalinovskiy/upf-cobalt.git
###Usage To use the metric, run evaluate.py with the following parameters:
| Parameter | Description |
|---|---|
| -r | parsed reference file |
| -t | parsed system output file |
| -v | file containing word vectors - optional |
| -a | output alignments - optional |
| -o | specify output directory - optional (default is "./Data") |
If no parameters are specified the metric will process the example files stored in Data folder.