Function main

nlp_class2/pos_baseline.py:171–221

()

    return Xtrain, Ytrain, Xtest, Ytest, word2idx

def main():
    Xtrain, Ytrain, Xtest, Ytest, word2idx = get_data()

    # convert to numpy arrays
    Xtrain = np.array(Xtrain)
    Ytrain = np.array(Ytrain)

    # convert Xtrain to indicator matrix
    N = len(Xtrain)
    V = len(word2idx) + 1
    print("vocabulary size:", V)
    # Xtrain_indicator = np.zeros((N, V))
    # Xtrain_indicator[np.arange(N), Xtrain] = 1

    # decision tree
    dt = DecisionTreeClassifier()

    # without indicator
    dt.fit(Xtrain.reshape(N, 1), Ytrain)
    print("dt train score:", dt.score(Xtrain.reshape(N, 1), Ytrain))
    p = dt.predict(Xtrain.reshape(N, 1))
    print("dt train f1:", f1_score(Ytrain, p, average=None).mean())

    # with indicator -- too slow!!
    # dt.fit(Xtrain_indicator, Ytrain)
    # print("dt score:", dt.score(Xtrain_indicator, Ytrain))

    # train and score
    model = LogisticRegression()
    model.fit(Xtrain, Ytrain, V=V)
    print("training complete")
    print("lr train score:", model.score(Xtrain, Ytrain))
    print("lr train f1:", model.f1_score(Xtrain, Ytrain))


    Ntest = len(Xtest)
    Xtest = np.array(Xtest)
    Ytest = np.array(Ytest)
    # convert Xtest to indicator
    # Xtest_indicator = np.zeros((Ntest, V))
    # Xtest_indicator[np.arange(Ntest), Xtest] = 1

    # decision tree test score
    print("dt test score:", dt.score(Xtest.reshape(Ntest, 1), Ytest))
    p = dt.predict(Xtest.reshape(Ntest, 1))
    print("dt test f1:", f1_score(Ytest, p, average=None).mean())
    # print("dt test score:", dt.score(Xtest_indicator, Ytest)) # too slow!

    # logistic test score -- too slow!!
    print("lr test score:", model.score(Xtest, Ytest))
    print("lr test f1:", model.f1_score(Xtest, Ytest))

if __name__ == '__main__':
    main()