false_etymologies.py

#!/usr/bin/env python3

"""Attempt to generate false etymologies from phonetically similar words"""

import csv
import json
import sys
import random

from itertools import groupby

from metaphone.metaphone import doublemetaphone

POS = {
     'adj.',
     'adv.',
     'conj.',
     'interj.',
     'n.',
     'nj.',
     'pl.',
     'prep.',
     'pron.',
     'v.'
}

def double_metaphones(words):
    grouped = groupby(sorted(words, key=doublemetaphone), doublemetaphone)
    return {dm: list(words) for (dm, words) in grouped}

def query(word, pos, etyms):
    # this potentially throws some results away...
    results = [
        e for e in etyms.get(word) if 
        (pos is None) or (e.get('pos') is None) or (e.get('pos') == pos)
    ]
    if results:
        result, *_ = results
        return result

def candidates(word, pos, etyms, dms):
    # the first is the real etymology
    real = query(word, pos, etyms)
    if real is not None:
        yield real
    # generate fake etymologies
    for similar in dms[doublemetaphone(word)]:
        for sim_etym in etyms[similar]:
            yield sim_etym

if __name__ == '__main__':
    import argparse
    parser = argparse.ArgumentParser(
        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
        description=__doc__
    )
    parser.add_argument(
        'query',
        help=(
            'a word to query for its true etymology and candidate false '
            'etymologies'
        )
    )
    parser.add_argument(
        '-p', '--pos',
        default=None,
        choices=POS,
        help='help'
    )
    parser.add_argument(
        '-e', '--etymologies',
        type=argparse.FileType('r'),
        default='etymologies.json',
        help=(
            'a JSON file with etymological entries keyed on word strings '
            '(e.g., https://www.dropbox.com/s/mlik0dtqv9pf72v/etymology.json)'
        )
    )
    parser.add_argument(
        '-n',
        type=int,
        default=None,
        help=(
            'limit results to the first n '
            '(all results are returned by default)'
        )
    )
    parser.add_argument(
        '-s', '--seed',
        default=None,
        help='seed string for random initialization',
    )
    args = parser.parse_args()
    writer = csv.writer(sys.stdout, delimiter='\t')
    # echo the query
    writer.writerow((args.query, args.pos))
    if args.seed is not None:
        random.seed(args.seed)
    # compile etymologies dict
    etyms = json.loads(args.etymologies.read())
    # compile double metaphone dict
    dms = double_metaphones(etyms.keys())
    # get real and fake etymology candidates
    real, *fakes = candidates(args.query, args.pos, etyms, dms)
    etymologies = list({f['definition'] for f in fakes})
    if len(etymologies) <= 1:
        print('Not enough etymologies!', file=sys.stderr)
        sys.exit(1)
    if args.n is not None and args.n <= len(etymologies):
        etymologies = random.sample(etymologies, k=args.n - 1)
    etymologies.append(real['definition'])
    random.shuffle(etymologies)
    for i, etymology in enumerate(etymologies, 1):
        writer.writerow((i, etymology))

Requires Metaphone:

pip3 install Metaphone