BabelNet API

As well as using WordNets for synonym look-up, we can use the BabelNet API to generate lists of hyponyms, cohyponyms and hypernyms.

The BabelNet look-up script can be started with the following command:

java -Xmx70g -cp tracer.jar eu.etrap.tracer.preprocessing.external.semantic.BabelNetTestMain ../lemma-list.txt EN >babelnet-queries.log

Where:

• -Xmx70g defines the memory to be used on the server; with a 70 gigabyte memory the BabelNet API look-up can take up to three days to compute.

• lemma-list.txt (a better file name might be lemma-babelnet.txt) is a modified version of TRACER's lemma.txt file containing only the base-form of every word in the text and its PoS tag (so, a two column file separated by a TAB). Place the lemma-list.txt in the TRACER's data/corpora/... directory and define its path in the command above.

• babelnet-queries.log is the log.

The script crunches two parameters: the lemma-list.txt and the language code (e.g. EN if you are working with English texts, DE if you are working with German texts, etc.). All generated files will be placed in the TRACER directory.

With cat babelnet-queries.log | grep "PROCESSING WORD" | wc -l you see the total number of processed words out of the total number of base-forms in lemma-list.txt.