A Study On The Searching Decision-making Learning Behavior Of Academic Database Users

The problem this paper studied can be retrieved from the necessary of the retrieval lessons in university. As we all know, the goal of the retrieval lesson which had spreaded several years ago is to help college students improve their academic knowledge. On the other hand, it is also an important way to service users and improve the efficiency of resource utilization.With the development of Internet, the function of information system is more and more easy to use. So many people think it is unnecessary to learn searching knowledge. However, the data which we have investigated shows that not only the freshmen, but also senior students, master students, doctoral students cannot use many basic searching methods. In fact, when surfing Internet, many people donot have specific targets. While using academic database, the targets are more specified. So there is a learning problem as for using academic database.Further, we found that the new generation which relies on Internet deeply prefers on-line help comparing with off-line training. So it is important to provide more efficient interactive network environment. Then it is more and more important to investigate the searching mechanism of academic user in the interactive environment, and it is the reason why this paper proposes the issue.Based on the analysis, this paper present the following academic problem:what is the learning mechanism when the academic users who have behavior habits learn a new searching resource? And whether the learning mechanism would change when under different conditions, conclding personal factors and environmental factors?The learning content this paper mentioned is learning how to choose proper seaming methods and keywords input mode. In other words, it is the learning problem about information retrieval decision and it is also the help which would afforded by the system. Learning is a process of acting adjustment, which is stimulated by result and leaded to change perceptions. For example, a user decides whether he should contiue use the same seaming method or change to another one according to the last seaming result.The learning problem is widely studied in the field of information retrieval. And those researchs which concern similar issues mainly about the experiment analysis of control groups. While this paper observes the characters during the changing process by holding multi-turn experiment. And we mainly use learning models to reveal and describe the learning rules.The main work of this paper is as following.(1) This paper organizes and analyses related theory and methods. Mainly, we summarize the related learning theory and learning models.(2) This paper analyse the learning behavior and its mechanism during the decision-making decision process. the process which including choosing search tools, accessing search results, results feedback and adjusting search tools could be abstracted as a cognize evolutionary and behavior adjustment. Based on the anylsis, we propose the hypothesis that there would be reinforcement learning and observation learning phenomon, as well as there would be different learning characters in different cognitive conditions and intervention environment.(3) Analysing the application of learning models, it would provide the improtant method to analysis the decision-making learning behavior. This paper studies the learning characters and the appropriate for using reinforcement learning models and fifctious learning models. And we modify the fifcitious model according to the characoters of the academic user.(4) Experimental study. This part is the most important part of this paper, and it is also embodied the innovation and difficulty. The main purpose is using quasi-experiment to test the hypothesis which is proposed in the theory analysis part by using learning models as method.The design idea of quasi-experiment we proposed could not only overcome the interference factor from the traditional control experiment, but also avoid the suspicious value of abstract experiment which usually used in game. We design the uniformity of standards of understanding to control the interfering of different cognition to the need and the evaluation of the results.We also construct the artificial environment to investigate the "observing experiment" in order to the effect of example information based on conesulting the game experiment design and summarizing users'behavior characters. According to the results, we achieve the desired effect.Furthermore, we reveal the characters of decision-making behavioer adjustment during the process of reinforcement learning and observing learning by using learning models, according to the models fitting investigation methods, we design the parameters of initial value, strategies set. study preference and stability which is different from the play. Then we achieve the aim that using learning models to research the characters of reinforcement learning and observation learning. According to the results, we achieve the expected targets. The experiment results indicate that the experiment design is feasible and the results are meaningful. According to the results, we not only verify the hypothesis, but also acquire the specific information about the learning behavior. In particular, the modified models have got satisfied fitting performance.The reinforcement learnig experiment results indicate that:the learning behavior is according to the "result rule" as for both freshmen and senior students. Specificly. the Markov character is more obvious as for freshmen. The character as for the senior students is according to the experience rules and the character is more obvious as for independent field users than dependent users.The observing learning experiment results indicate that:the observation character is obvious and there is also inforcement learning characters during the process. And the modification of fifctious learning models is effective. The fitting results are much better for dependent field users than for independent field users.(5)Conclsion. Based on the results of model fitting and the data from video and questionnaire, we analyse the external behavior and internal mental characters of the "reinforcement learning" and "observing learning" according to the related theories. And then we discuss the possibility of construction the future online help system and offline training service.