e33d1ba9150a3084947ae000f12a9b6f2f644648,conceptnet5/vectors/evaluation/wordsim.py,,evaluate_semeval_monolingual,#,307

Before Change

    scores = []
    for lang in ["en", "de", "es", "it", "fa"]:
        scores.append(spearman_evaluate(vectors, read_semeval_monolingual(lang)))
    top_scores = sorted(scores, key=lambda x: x["acc"] if not np.isnan(x["acc"]) else 0)[-4:]
    acc_average = tmean([score["acc"] for score in top_scores])
    low_average = tmean([score["low"] for score in top_scores])
    high_average = tmean([score["high"] for score in top_scores])
    return pd.Series(
        [acc_average, low_average, high_average],
        index=["acc", "low", "high"]
    )

def evaluate_semeval_crosslingual(vectors):
    
    According to Semeval2017-Subtask2 rules. the global score is the average of the individual

After Change

    
    Get a semeval score for a single monolingual test set.
    
    spearman_score = measure_correlation(spearmanr, vectors, read_semeval_monoling(lang))
    pearson_score = measure_correlation(pearsonr, vectors, read_semeval_monoling(lang))
    score = compute_semeval_score(spearman_score, pearson_score)
    return score


def evaluate_semeval_crosslingual(vectors, lang1, lang2):
    

Instances