Abstarct: Privacy policies are the primary channel through which companies inform users about their data collection and sharing practices. In their current form, policies remain long and difficult to comprehend, thus merely serving the goal of legally protecting the companies. Short notice baxsed on information extracted from privacy policies have been shown to be useful and more usable. Therefore, we deal with new information retrieval systems to meet this need. Question answering systems are examples of such ones which use natural language to make human-computer interaction much easier. Performance of statistical question answering systems are so dependent to their data-set with which they are trained; however, they outperform other approach of implementation of question answering. Deep convolutional neural network equipped with interaction will consider previous question and similarity of (past and new questions) questions while answering client. Thus, they omit majority of ambiguity errors. Improvement are resulted by word2vec model which can identify similarity of meaning of two sentences. The result of this research implementation can achieve an accuracy of %82.4 on this task, when evaluated against unseen test set annotated by a group of law department student. The obtained scores indicate that the proposed method is almost %8 more accurate than automated analysis of privacy policy using only deep learning. In addition, the processing time of the proposed method compared with the hierarchy of neural network classifiers is much improved.