1. Introduction to text mining 2. Core text mining operations 3. Text mining preprocessing techniques 4. Categorization 5. Clustering 6. Information extraction 7. Probabilistic models for Information extraction 8. Preprocessing applications using probabilistic and hybrid approaches 9. Presentation-layer considerations for browsing and query refinement 10. Visualization approaches 11. Link analysis 12. Text mining applications Appendix Bibliography.

/pdf/the-text-mining-handbook-advanced-approaches-in-analyzing-ge55wa29fm.pdf

The Text Mining Handbook: Advanced Approaches in Analyzing Unstructured Data

With today’s large increase in digital images and automatically generated imagery, such as videos and stills generated from surveillance equipment, the need for efficient image retrieval and indexing has become fundamental. Since text-based information retrieval has been shown to perform very poorly when searching through images, research has been active in the field of content-based image retrieval (CBIR). CBIR systems make use of the properties of images in order to compare them and extract content by matching the query image. Comparing features – such as color, texture, and shape – allows for better retrieval accuracy; however, the algorithms used are still very limited. This paper will provide a survey of CBIR systems and explain the fundamental properties and techniques used in these systems. First, the history of CBIR systems will be discussed together with some typical CBIR systems. After this, the paper will touch on text-based information retrieval and explain why it does not work for searching through collections of images. The latter portion of this document will provides an overview of a typical CBIR system and the main techniques involved in querying such a system. Finally, image features and indexing schemes will be described.

Content-Based Image Retrieval

As the number of Internet users and the number of accessible Web pages grows, it is becoming increasingly difficult for users to find documents that are relevant to their particular needs. Users must either browse through a large hierarchy of concepts to find the information for which they are looking or submit a query to a publicly available search engine and wade through hundreds of results, most of them irrelevant. The core of the problem is that whether the user is browsing or searching, whether they are an eighth grade student or a Nobel prize winner, the identical information is selected and it is presented the same way. In this paper, we report on research that adapts information navigation based on a user profile structured as a weighted concept hierarchy. A user may create his or her own concept hierarchy and use them for browsing Web sites. Or, the user profile may be created from a reference ontology by 'watching over the user's shoulder' while they browse. We show that these automatically created profiles reflect the user's interests quite well and they are able to produce moderate improvements when applied to search results.

Ontology-based personalized search and browsing

Data Clustering: Theory, Algorithms, and Applications (ASA-SIAM Series on Statistics and Applied Probability)

With the advent of the Internet, we are seeing more sophisticated techniques being developed to support e-Iearning. The rapid developme nt of Web-based learning and new concepts like virtual classrooms, virtual laboratories and virtual universities introduces many new issues to be addressed. On the technical side, we need to develop effective e-technologies for supporting distance education. On the learning and management side, we need to consider issues such as new style of learning and different system set-u p requirements. Finally, the issue of standardization of e-Iearning systems should also be considered. In this special issue, our focus will be on the technical side, although other issues related to knowledge and data engineering for e-Iearning may also be considered. Topics: In this special issue, we call for original papers describing novel knowledge and data engineering techniques that support e-Iearning. Preference will be given to papers that include an evaluation of users' experience in using the proposed methods. Areas of interests include, but are not limited to: • Semantic Web technology for e-Iearning • Data modeling (eg., XML) for efficient management of course materials • Searching and indexing techniques to suppo rt effective course notes retrieval • User-centric e-Iearning systems and user interaction management • Profiling techniques to support grading and learning recommendation • Data and knowledge base suppo rt for pervasive e-Iearning • Course material analysis and understanding • Automatic generation of questions and answers • Collaborative communities for e-Iearning

Knowledge and Data Engineering for e-Learning Special Issue of IEEE Transactions on Knowledge and Data Engineering

Designing efficient and fault-tolerant algorithms for solving the mutual exclusion problem is an important issue in distributed systems. Many algorithms have been proposed in the literature, and they are being broadly classified into two types: token-based and permissionbased [5, 21]. In a token-based algorithm, the system uses a unique token to represent the privilege for the node possessing it to enter the critical section. In a permissionbased algorithm, a node wishing to enter the critical section needs to request permissions from a subset of other nodes, called a request set. Earlier results typified by [14, 19] allow the nodes in the request sets to grant multiple permissions, but use logical clocks and timestamps to ensure mutual exclusion and freedom of deadlocks. A different approach typified by [16] requires that each node in a request set grant one permission at a time. Such a request set is called a quorum, and the family of all quorums forms a coterie if certain minimality condition is satisfied. Quorum-based algorithms guarantee mutual exclusion because each pair of the quorums have a non-empty intersection. The main advantage of quorum-based algorithms is their resilience to node and network failures, because any quorum with fully operational nodes can grant permissions, improving the system’s availability. In this paper, we are mainly interested in quorum-based algorithms for solving the distributed mutual exclusion problem, although similar techniques have been applied to solving the replicated data control problem [12, Chap.7].

A Torus Quorum Protocol for Distributed Mutual Exclusion

We describe a natural extension of the banker's algorithm (D.W. Dijkstra, 1968) for deadlock avoidance in operating systems. Representing the control flow of each process as a rooted tree of nodes corresponding to resource requests and releases, we propose a quadratic-time algorithm which decomposes each flow graph into a nested family of regions, such that all allocated resources are released before the control leaves a region. Also, information on the maximum resource claims for each of the regions can be extracted prior to process execution. By inserting operating system calls when entering a new region for each process at runtime, and applying the original banker's algorithm for deadlock avoidance, this method has the potential to achieve better resource utilization because information on the "localized approximate maximum claims" is used for testing system safety.

An extended banker's algorithm for deadlock avoidance

To enable efficient similarity search in large databases, many indexing techniques use a linear transformation scheme to reduce dimensions and allow fast approximation. In this reduction approach the approximation is unbounded, so that the approximation volume extends across the dataspace. This causes over-estimation of retrieval sets and impairs performance.This paper presents a non-linear transformation scheme that extracts two important parameters specifying the data. We prove that these parameters correspond to a bounded volume around the search sphere, irrespective of dimensionality. We use a special workspace-mapping mechanism to derive tight bounds for the parameters and to prove further results, as well as highlighting insights into the problems and our proposed solutions. We formulate a measure that lower-bounds the Euclidean distance, and discuss the implementation of the technique upon a popular index structure. Extensive experiments confirm the superiority of this technique over recent state-of-the-art schemes.

/pdf/a-non-linear-dimensionality-reduction-technique-for-fast-hxpg8n2hg9.pdf

A non-linear dimensionality-reduction technique for fast similarity search in large databases

We propose several algorithms using the vector space model to classify the news articles posted on the NETNEWS according to the newsgroup categories. The baseline method combines the terms of all the articles of each newsgroup in the training set to represent the newsgroups as single vectors. After training, the incoming news articles are classified based on their similarity to the existing newsgroup categories. We propose to use the following techniques to improve the classification performance of the baseline method: (1) use routing (classification) accuracy and the similarity values to refine the training set; (2) update the underlying term structures periodically during testing; and (3) apply k-means clustering to partition the newsgroup articles and represent each newsgroup by k vectors. Our test collection consists of the real news articles and the 519 subnewsgroups under the REC newsgroup of NETNEWS in a period of 3 months. Our experimental results demonstrate that the technique of refining the training set reduces from one-third to two-thirds of the storage. The technique of periodical updates improves the routing accuracy ranging from 20% to 100% but incurs runtime overhead. Finally, representing each newsgroup by k vectors (with k = 2 or 3) using clustering yields the most significant improvement in routing accuracy, ranging from 60% to 100%, while causing only slightly higher storage requirements.

Classification algorithms for NETNEWS articles

Distance-based outlier detection is an important data mining technique that finds abnormal data objects according to some distance function. However, when this technique is applied to high-dimensional categorical data, a traditional simple matching dissimilarity measure does not provide an adequate model. In this article, we employ a new common- neighbor-based distance function to measure the proximity between a pair of data points. Experiments show that better outlier mining results can be achieved when the new distance function is utilized rather than a conventional simple matching dissimilarity measure.

Sheau-Dong Lang

Papers

A Torus Quorum Protocol for Distributed Mutual Exclusion

An extended banker's algorithm for deadlock avoidance

A non-linear dimensionality-reduction technique for fast similarity search in large databases

Classification algorithms for NETNEWS articles

Mining Distance-Based Outliers from Categorical Data