Dissertation > Excellent graduate degree dissertation topics show

A New Annotate Ontology Method Based on Bootstrapping

Author: GaoZuo
Tutor: LuoJun
School: Chongqing University
Course: Computer System Architecture
Keywords: Weak supervision Rule Noumenon Marked
CLC: TP391.1
Type: Master's thesis
Year: 2010
Downloads: 97
Quote: 0
Read: Download Dissertation

Abstract

With the development of the Internet, web resources showing rapid growth, but low Internet information processing automation, the correlation between the information poor, even with a powerful search engine, due to excessive redundant information, and can not quickly and accurately web resources for effective information. To solve this problem, the Web founder Tim Berners-Lee in 1998, the concept of the Semantic Web, it expanded layer is added on top of the existing Web infrastructure, and formal description of Web information in this layer . Marked by bulk vocabulary Web resources, the state of the resources on the Web from the machine readable to machine-understandable degree of Semantic Web-based development and efficient access Web information solution. Most of the tagging method low degree of automation, poor adaptability and inefficient. Ontology annotation method, carried out a systematic study to explore a new method based Bootstrapping ontology annotation. First resolve a given ontology, to generate the rules file, and then filter out the field of document text classification. After using Bootstrapping label extraction and Ontology reasoning, after a few cycles, using only a small amount of training text can achieve better annotation. The main work of this paper are as follows: ① proposed a new automatic learning algorithm based Bootstrapping and Bayesian text classification algorithm. Often complex and diverse to be marked text, if information directly labeling, extraction, a huge workload marked the high error rate. Before labeling text classification related to the field of Ontology extracted documentation. To make the classification correct classification and labeling in the case of small sample text, this paper presents a new automatic learning algorithm based Bootstrapping and Bayesian text classification algorithm, the algorithm only need a small amount of training samples as seed set to train the classifier, and then from the results of the classification of the selected part of the highest confidence text added to the seed concentration, was repeated as a new round of training samples, the training until the end. In this way, only by a small amount of training samples will be able to reach a large number of training samples training results. ② Bootstrapping and rule text set marked A. First of all, according to the rules file initial annotation of text set, marked a good set of texts. Then the instance of the context of the relationship, drawing the WHISK algorithm of extraction rules induction, generate new rules file, marked a new vocabulary. Subsequently, the extraction of the information label, filled into the bulk file. Finally, by means of the Ontology Inference Engine, reasoning ontology file, remove erroneous data, and trim the wrong rules, After several iterations, so that the model can achieve automatic extraction of the new instance, enrich, improve the body's purpose. The pending iteration is complete, marked good text collection and a rich ontology library. The ③ proposed method based Bootstrapping ontology annotation. The field of text classification and labeling combined into an overall model, model after each iteration expanded ontology library makes the classifier to continue, but also further expand the ontology the unlabeled field documentation generated by the classification. The cycle is repeated, the reach of the small sample training set ontology annotation purposes. After a large number of experiments show that this method has better classification results with higher precision and recall rate of the ontology annotation.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Text Processing
© 2012 www.DissertationTopic.Net Mobile