Research On Intelligent Human-Computer Interaction Methods In Information Retrieval Systems

Along with the rapid development of the Internet, the scale of web information is becoming larger and larger. Web becomes an important knowledge repository; it is highly desirable for people to obtain wanted information efficiently. Through continuous development and improvement, information retrieval has gained more and more attention. Existing information retrieval system still exist some problems and defects, such as information organization, and other intelligence interactive issues. Those problems have received wide spread attention in both academia and industry.This paper mainly focuses on two information retrieval sub-systems’intelligence interactive methods:(1) image retrieval system search result visualization,(2) intelligent dialog management strategy in conversational spoken system. To enhance the intelligence and interactiveness in retrieval system, the former one changes the organization and management of information, the latter changes the mode to obtain information.1. Image retrieval result visualization. Most of existing popular search engines (e.g., Google, Baidu) allow users to represent their search intents by issuing the query as a list of keywords. However, there are several defects in these search engines. Firstly Keyword queries are usually ambiguous especially when they are short. Secondly, the result is displayed as an image list which cannot provide an overview of whole image result intuitively. In order to find target images, users have to scan the whole result list repetitively. Facing this situation, we propose a novel tag cloud scheme named Visual Tag Clouds (VTC), VTC provides not only textual tags, but also image examples correlated with tags in the tag clouds. This paper seeks to utilize VTC to provide a visual summarization for query result, which is more intuitive than image list.VTC is created in two stages, tag selection and image selection. In the tag selection stage, we present a series of strategies for ranking and selecting tags. We incorporate a variety of similarities among tags into frequency-based methods, including co-occurrence, semantic and visual similarity among tags respectively as well as the combination of them. In the image selection stage, we employ the Affinity Propagation (AP) algorithm to select exemplar images for each tag selected in the first stage. The experiment result shows that VTC is able to specify users’ query intentions more precisely as well as summarize and navigate query results efficiently.2. Intelligent dialog management strategy in conversational spoken system. Based on Agenda-based method, by introducing user information agent (UIA), we achieved an information-independent, task-independent dialog management framework. This framework isolates the domain-specific aspects of the dialog control logic from domain-independent conversational skills, and in the process facilitates rapid development of mixed-initiative systems operating in complex, task-oriented domains.The dialog task specification describes a hierarchical plan for the interaction. UIA maintains a history of the discourse, the dialog engine use it to interpret the perceived semantic inputs in the current context. The current semantic input, together with the current dialog state and information about the task to be performed is then used to determine the next system action. In order to construct multi-task spoken system, we designed a control-sub-tree-based keyword triggered theme transfer strategy. By mining user interaction behaviors, on the basis of multi-Markov decision processes, this paper designed a four-step error handling model (4-MDP), for helping users to rectify errors. The experiment result shows that methods we proposed are more effective.