evaluate_names_matching.py

from tqdm.notebook import tqdm

results = []
for norm_name, normalize in tqdm((("simple", normalize_simple),
                                  ("unidecode", normalize_unidecode),
                                  ("unidecode_metaphone", normalize_unidecode_metaphone))):
    for comp_name, compare in tqdm((("ratio", distance_ratio),
                                    ("sort_ratio", distance_sort_ratio),
                                    ("set_ratio", distance_set_ratio),
                                    ("ratio_join", distance_ratio_join),
                                    ("max_ratio", distance_max_ratio))):
        for match_name, match in tqdm((("greedy", greedy_match),
                                       ("lap", lap_match))):
            def evaluate(names1, names2):
                series1, series2 = (
                    [normalize(" ".join(sorted(strs))) for strs in names]
                    for names in (names1.values(), names2.values())
                )
                matches = match(series1, series2, compare)
                keys2 = list(names2)
                truth = [keys2.index(key) for key in names1]
                return (matches == truth).sum() / len(truth)
            
            results.append((np.mean([evaluate(*org) for org in dataset.values()]),
                            norm_name, comp_name, match_name))