如何使用线程减少执行时间？

Question

我有一个程序找到单词对的Pointwise互信息。我所做的代码如下：

temp_list = co_occurrence_dict.values()
N=0
for item in temp_list:
N += item

    main_dict = {}
    temp_dict = {}
    for word in all_opinion_words:
        for key,value in co_occurrence_dict.iteritems():
            if ((word == key.split()[0]) or (word == key.split()[1])):
                pmi_eqn_numerator = (value)/float(N)
                if (key.split()[0] != word):
                    temp_word = key.split()[0]
                else:
                    temp_word = key.split()[1]
                if temp_word==word:
                    temp_dict[temp_word] = 0
                else:
                    temp_sum = generate_sum_occurrence_values(temp_word,co_occurrence_dict)
                    pmi_eqn_denominator1 = (temp_sum)/float(N)
                    temp_sum = generate_sum_occurrence_values(word, co_occurrence_dict)
                    pmi_eqn_denominator2 = (temp_sum)/float(N)
                    pmi = math.log(pmi_eqn_numerator / (float(pmi_eqn_denominator1) * float(pmi_eqn_denominator2)))
                    temp_dict[temp_word] = pmi
        temp_list = []
        for elements in temp_dict.iteritems():
            temp_list.append(elements)
        main_dict[word] = temp_list

def generate_sum_occurrence_values(item,temp_dict):
    sum_values = 0
    for key,value in temp_dict.iteritems():
        if ((item == key.split()[0]) or (item == key.split()[1])):
            sum_values += value
    return sum_values

其中，all_opinion_words是单词列表，co_occurrence_dict的形式为

co_occurrence_dict={'social contemporary': 1, 
                    'earthly indeed': 1,
                    'far mythical': 1,
                    'small higher': 1, 
                    'ideological even': 1, 
                    'certain `magnificent': 1,
                    'back al': 8,
                    'perhaps thin': 1,
                    'never skeptical': 1,
                    'federal small': 1,
                    'difficult most': 1, 
                    'also young': 1, 
                    'ideological ever': 1,
                    'far rather': 1,
                    'able happy': 1}

由于我必须处理大量文档，所以all_opinion_words和co_occurrence_dict的大小非常大。因此需要很长的执行时间。我读过线程可以用来加速执行。如何在此代码中应用线程？