Quering metaphors

1. Cd to the repository root:

cd /lfs1/vzaytsev/software/metaphor-search/

2. Add metaphor's packages to PYTHONPATH and set NLTK data path (if needed):

HUGIN_ROOT=$PWD/hugin
SEAR_ROOT=$PWD/sear
METAPHOR_ROOT=$PWD/metaphor

export PYTHONPATH=PYTHONPATH:$HUGIN_ROOT:$SEAR_ROOT:$METAPHOR_ROOT
export NLTK_DATA=/lfs1/vzaytsev/nltk

3. Set searcher options:

# Specify language, choices: ru, es, en
LANG=en
SENTENCES_ROOT=/lfs1/vzaytsev/corpora2/$LANG/sentence_index
DOCUMENTS_ROOT=/lfs1/vzaytsev/corpora2/$LANG/document_index
# change this according to your query file:
QUERIES=/lfs1/vzaytsev/corpora2/$LANG/queries/query.*.json # this can be a pattern
# set any location for found metaphors
OUT_DIR=/lfs1/vzaytsev/corpora2/$LANG/found/

4. Run searcher:

python scripts/run_sentence_candidates.py \
		-i $SENTENCES_ROOT \
		-c $DOCUMENTS_ROOT \
		-l $LANG \
		-f json \
		-q $QUERIES \
		-o $OUT_DIR \
		-x 0 \
		-e ".metaphors" \
		-p 1

5. Options:

1. -i, --input LF sentences index root location.
2. -c, --context_input Context index root location.
3. -l, --language Data language: ru, es, or en.
4. -f, --format Output format: json or plain.
5. -o, --output Location of the output directory.
6. -x, --output_lf Put logic form of found metaphor to output file: 0 or 1.
7. -q, --query Metaphor search query file.
8. -e, --extension Extension that will be addet to the output files. Default is .metaphors.json.
9. -p, --use_pos Use parts of speech. Postfixes such as -\w+ (-n, -vb, etc) if present, will be recognised as additional part of speech constraint in query. Default is 0.

Query Example

{
    "annotation": {
        "label": "poverty",
        "language": "english",
        "corpus": "gigaword",

        "source_frame": "building",
        "source_concept_subdomain": "",

        "target_frame": "wealth",
        "target_concept_domain": "economy",
        "target_concept_subdomain": "wealth"
    },
    "query": {
        "max_path_lenght": 100,
        "targets": [
            "money-n"
        ],
        "sources": [
            "water-n"
        ]
    }
}