Dissertation > Excellent graduate degree dissertation topics show

Research on Markov Graph Model in Information Retrieval

Author: ZuoJiaLi
Tutor: WangMingWen
School: Jiangxi University of Finance
Course: Management Science and Engineering
Keywords: Information Retrieval Model Text Classification Model Markov Network Term Relationship Term Importance
CLC: TP391.3
Type: PhD thesis
Year: 2011
Downloads: 57
Quote: 0
Read: Download Dissertation

Abstract


With the rapid development of the Internet and the process of globalization, web information resources increase exponentially. Large-scale information has changed the traditional way of knowledge acquisition, making the Internet has become a major source of knowledge. How to use these large, heterogeneous, unstructured information resources has become an important issue and requires an urgent solution. Information retrieval technology as a key technology to meet this challenge gets a large number of concerns and significant development.Traditional information retrieval models have been used in many areas of applications, and achieved considerable success, but current information retrieval model is still facing many difficulties and has no good solution. With the continuous development of the Internet and increasingly high demand for information retrieval, how to construct information retrieval model with better performance will be a focused research topic.Based on random process theory, we use Markov network to model term relationship, which different from the traditional way, to develop a general information retrieval model and text classification model in a unified framework. The main contributions include:(1) Information retrieval model based on Markov network representationIn information retrieval, the relevance of queries and document can be measured by the importance of the term. But most of the models suppose that term is independent with each other, which is rarely true in reality. Actually, there is a strong correlation between terms. In addition, as most information retrieval models are based on queries, there are many researches about the analysis of queries and proposed a number of methods, such as query expansion, relevance feedback. The goal of information retrieval is to find relevant documents, which makes the analysis only to the query is not enough and also requires further analysis of the document.We use Markov network to model term relationship, as to get better document models, and develop information retrieval model in a unified framework which can model term relationship and information retrieval model to explore the impact of relevance information to retrieval performance. The Markov network representation model can model arbitrary features including term relationship and all kinds of term features. Based on the analysis of the complexity of the network structure, we explore some rules for information retrieval model construction, and construct the information retrieval model following this set of rules. Experimental results show that the relevance information can improve the retrieval model performance effectively.(2) Query expansion, document expansion and Feedback model based on Markov network representationAs the importance of terms in the retrieval is of the great impact on retrieval performance, according to which, we propose the idea of term importance to the information retrieval. Corresponding to these definitions, we given some other rules and combine query expansion, document expansion and feedback technique to construct information retrieval models. Our model uses the term relationship to filter this noisy information which can ensure the information retrieval performance of the model. Experimental results show that query expansion model and Feedback model show good performance.(3) Text Classification Model Based on Markov RepresentationTraditional text classification models are based on vector space model, which takes the features independent assumption and discard some useful information about text classification. In this paper, we propose Text Classification Model Based on Markov Network and Text classification model base on Markov Representation.Text classification based on Markov network model propose a new priori probability estimation method to address the a priori probability problem in Naive Bayes model at first, then model the term relationship in text classification model. Text classification model base on Markov Representation represente the document Markov networks, and classify a document based on distance between Markov networks. Experimental results show that both models can improve the classification performance.

Related Dissertations

  1. Face Image Super-Resolution Technology,TP391.41
  2. Markov Retrieval Model Based on Transferring Learning,TP391.3
  3. Extended Information Retrieval Model Based on Markov Cliques,TP391.3
  4. Research on Several Models in Text Classification and Clustering,TP391.1
  5. Kullback-Leibler distance framework abstracts retrieved in Retrieval,TP391.3
  6. Theme Extracting from Academic Paper Based on Domain-specified Concept System,TP391.1
  7. Study on the Diversity of Insect Community and the Seasonal Dynamics in the Neighborhood System of Paddy with Pomegranate,S186
  8. Face cartoon based on machine learning method,TP391.41
  9. Research on the Relationship Words between Qintang Goulou Cantonese Dialect and the Language of Zhuang,H218
  10. Research on Combining Collective Classification with Active Learning,TP181
  11. Extended SBN Retrieval Model Based on Ontology Terms Relationship,TP391.3
  12. Markov network model Human Face , image super-resolution algorithm,TP391.41
  13. Research on Learning-Based Nonlinear Algorithm of Face Image Super-Resolution,TP391.41
  14. Research on Example-Based Single Image Super-Resolution,TP391.41
  15. EDGE DETECTION BASED ON MACHINE LEARNING,TP391.41
  16. Information Retrieval Model Based on Markov Network,TP393.09
  17. The Research on Latent Semantic Markov Network Retrieval Model,TP391.3
  18. Network link transmission control stochastic Petri net modeling and performance analysis,TP393.02
  19. The Research on the Information Retrieval Based on the Markov Network,TP391.3
  20. Learning-Based Image Super-Resolution,TP391.41
  21. Research and Implementation of Two Text Categorization Algorithms,TP301.6

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