Research On Knowledge Mining In Person Tracking

There has been great improvement in the technology of information retrieval, textcategorization, text clustering and information extraction. Now the research has focused onsolving new problems using various methods. A series of evaluations has pushed forward thedevelopment of related research fields. In the field of mobile communication, the technology oftracking a person has been comparatively mature. But in the field of text mining, the method oftracking a person has not arisen enough attentions and has not drew a conclusion about it.Therefore, this article proposed the notion of Person Tracking. First it carried on theconcise introduction of Person Tracking, elaborated the connotation and the extension ofPerson and Person Tracking, then it proposed Person Model and carried on the formalizeddefinition to it. Afterwards, it produced the main question, the technical foundation, the systemflow as well as the application of Person Tracking.After making the summery of Person Tracking, this article introduced the method of datapreprocessing. The processed documents are mainly Chinese News WebPages. This articleresearched deeply of the method of HTML parsing and person attribute extracting. Throughpreprocessing, it turned the text content into fragment of Person Model.Then this article studied the identical judgement of Person Model. The main problem isabout processesing the fragments of Person Model, not only separating different person'sinformation, but also integrating information of the same person. This article used theinformation to make biography for the person. This article deeply introduced the featureexpressing and the similarity computing methods in this part.In the foundation of identical judgement of Person Model, this article studied the methodsof producing person's activity reports. This is so-called event-based information organization.The event organization divides into microscopic granularity event organization andmacroscopic granularity event organization. This article researched the macroscopic ones. Ittaked the standard international evaluation as the platform, taked the documents or theparagraphs as the basic processing units and did not involve the concrete event role frame. Inthis way, this article not only studied the macroscopic granularity event detection technology,but also studied the method of event relation organization.Afterwards, this article discussed the methods of several microscopic granularity tasks. Ittaked the circulation theory as the foundation, studied the method of the recognition degreecomputing. Through tracking a person's recognition degree in different periods, the person'srecognition charts can be drew.Through unifying the theory and the technical research of Person Tracking, this articlefinally proposed a method of buiding person information search engine. It not only designedthe system structure, but also produced the index structure and the interface of the searchengine.The research achievements of this article established solid foundation for the personinformation processing in free texts. The proposed Person Tracking provided a brand-newresearch angle for the field of text mining. And it had important practical value for the miningof abstract object.