def get_tfidf_idf_df_feature(collection_document_text, dict_dataset):
    doc_vocabulary = []
    list_doc = []
    order_doc_id = []
    for doc_id in dict_dataset:
        print(doc_id)
        doc_df = dict_dataset[doc_id]
        doc_vocabulary.append(np.array(doc_df.index.array).tolist())
        list_doc.append(collection_document_text[doc_id]) #Lista com os documentos. EM cada posição temos o documento  numa string
        order_doc_id.append(doc_id) # Ordem que vai ser utilizado pelo TFIDFVectorizer
    doc_vocabulary = [item for sublist in doc_vocabulary for item in sublist]
    doc_vocabulary = list(dict.fromkeys(doc_vocabulary))
    tfidf = TfidfVectorizer(vocabulary=doc_vocabulary,lowercase=True)
    tf = CountVectorizer(vocabulary=doc_vocabulary,lowercase=True)
    tf_matrix = tf.fit_transform(list_doc)

    tf_idf_matrix = tfidf.fit_transform(list_doc)

    feature_idf = dict(zip(tfidf.get_feature_names(), tfidf.idf_))
    df_idf = pd.DataFrame.from_dict(feature_idf, orient = "index")

    df_tfidf = pd.DataFrame(tf_idf_matrix.toarray(), columns=tfidf.get_feature_names())

    df_tf =pd.DataFrame(tf_matrix.toarray(), columns=tfidf.get_feature_names())

    for index,doc_id in enumerate(order_doc_id):
        token_document_list = dict_dataset.get(doc_id).index.tolist()

        df_tfidf_doc = df_tfidf.iloc[index]
        df_tf_doc = df_tf.iloc[index]
        df_idf_doc = df_idf.loc[token_document_list]

        if doc_id == '2':
            print("td",df_tf_doc["services"])
            print("tfidf", df_tfidf_doc.loc["services"])
            print("idf", df_idf_doc.loc["services"])

        df_filtered_tfidf = df_tfidf_doc[token_document_list]
        df_filtered_tf = df_tf_doc[token_document_list]
        dict_dataset.get(doc_id)["tfidf"] = df_filtered_tfidf
        dict_dataset.get(doc_id)["idf"] = df_idf_doc
        dict_dataset.get(doc_id)["tf"] = df_filtered_tf

    return dict_dataset