def learn1(data, unique_list):

    unique_list = []

    A = data[:, 3]
    train_labels = A[:len(A)/2]
    test_labels = A[len(A)/2:]
    '''
    B = unique_list
    train_labels = B[:len(B)/2]
    test_labels = B[len(B)/2:]
    '''
    C = data[:, 1]
    train_data = C[:len(C)/2]
    test_data = C[len(C)/2:]

    for i in train_data:
        if i == 'a':
            i = 0
        elif i == 'adv':
            i = 1
        elif i == 'infinitive-marker':
            i = 2
        elif i == 'det':
            i = 3
        elif i == 'n':
            i = 4
        elif i == 'pron':
            i = 5
        elif i == 'modal':
            i = 6
        elif i == 'v':
            i = 7
        elif i == 'conj':
            i = 8
        elif i == 'prep':
            i = 9
        elif i == 'interjection':
            i = 10
        unique_list.append(i)
    unique_arr = np.array(unique_list)
    print unique_arr
    #unique_labels = np.unique(labels)

    new_arr_list = []
    for i in train_data:
        temp_list = []
        new_arr_list.append(temp_list)
    new_arr = np.array(new_arr_list)

    clf = svm.SVC(kernel = 'linear')
    print clf.fit(new_arr, unique_arr)
    #return clf.predict(test_data)