Brain Cognitive Research On Topic Model

With the devolepment of current data mining technology,the topic model algorithm has been widely used.Using the topic model to analyze data have been more and more recognized by people.We know the theory of topic model algorithm,but we don’t know the cognitive processes of the human brain to think topic.We even don’t know whether the cognitive processes of the human brain thinking topic is consistent with topic model algorithm,this problem has not yet been explored.Modern,functional magnetic resonance imaging(fMRI),making it easier for people to get the data in each state of the brain,which is the preferred tool for interpretation of the human brain.In this paper,we use magnetic resonance imaging technology to explore the process of thinking topic of the human brain,and explore the differences between the subject model algorithm and the cognitive process of hunman brain.Specific inquiry work is as follows:(1)Locating the activate areas of the brain related to thinking topic about by an block-related fMRI experiment.Experimental results show that when the brain thinking topics of article,bilateral inferior parietal lobule,paracentral lobule,limbic,and cingulate cortex is activated.(2)Selecting feature voxels from the brain image to distinguish cognitive models of different types of articles.Using a variety of feature selection methods to extract feature from brian pictures,classification accuracy rate is over 70%,which proves cognitive model of the brain thinking about topic of different articles can be distinguished.(3)Finding the differences and consistences between topic model algorithm(LDA)and the cognitive process of thinking topics by human brain.The results demonstrated that there are differences on the level of the sentence,the human brain involved in thinking about sentence analysis process,and the topic model algorithm does not involve the sentence analysis.