Manabu Ohta

An Intelligent Web Recommendation System: A Web Usage Mining Approach

As an increasing number of Web sites consist of an increasing number of pages, it is more difficult for the users to rapidly reach their own target pages. So the intelligent systems supporting the users in navigation of the Web contents are in high demand. In this paper, we describe an intelligent recommendation system called the system L-R, which constructs user models by mining the Web access logs and recommends the relevant pages to the users based both on the user models and the Web contents. We have evaluated the prototype system and have obtained the positive effects.

Document warehousing: a document-intensive application of a multimedia database

Nowadays, structured data such as sales are stored in data warehouses for decision-making. Less-structured data such as HTML texts, XML data, images, and videos are increasingly accumulated in PC storage due to the spread of the Internet technology such as WWW. Such less-structured data, collectively called multimedia documents, are also precious as corporate assets. So we need to provide a document warehouse to analyze and manage multimedia documents for corporate-wide information mining and reuse like a data warehouse. As a document-intensive application of a multimedia database, we describe a prototype document warehouse system, which supports management of documents, keyword-based and content-based retrieval, rule-based classification, SOM-based clustering and XML active query facility based on ECA rules.

Web usage mining approaches to page recommendation and restructuring

As an increasing number of Web sites such as e-businesses consist of an increasing number of pages, users find it more difficult to rapidly reach their own target pages. Ill-structured design of Web sites also prevents the users from rapidly accessing the target pages. In this paper, we describe two complementary approaches to Web usage mining as a key solution to these issues. First, we describe an adaptable recommendation system called the L-R system, which constructs user models by classifying the Web access logs and by extracting access patterns based on the transition probability of page accesses and recommends the relevant pages to the users based on both the user models and the Web structures. We have evaluated the prototype system and have obtained the positive effects. Second, we describe another approach to constructing user models, which clusters Web access logs based on access patterns. The user models also help to discover unexpected access paths corresponding to ill-formed Web site design. Copyright

Project Xanadu: XML- and active-database-unified approach to distributed e-commerce

To model and implement EC business models like e-brokers on the Web, we need business modeling, distributed processing, and data compression as enabling technologies. We have designed an active query language for Web-distributed XML databases, called XBML, as an approach to business process modeling. We efficiently process XBML queries by using min-max and semi-join methods as distributed query optimization in the Web context. We also compress XML data for reducing overheads of storage and transmission and guaranteeing transparent access to compressed data through existing application interfaces. The authors describe the Xanadu project consisting of three research topics.

Active Knowledge Mining for Intelligent Web Page Management

In this paper, we describe active knowledge mining approaches to intelligent Web page management. Through application of them to operational Web systems, we have found them very effective as follows: First, we describe an adaptable recommendation system called the system L-R, which constructs user models as knowledge by classifying the Web access logs and by extracting access patterns based on the transition probability of page accesses and recommends the relevant pages to the users based both on the user models and the Web site structures. We have evaluated the prototype system and have successfully obtained the positive effects of the mined knowledge. Second, we describe another approach to constructing user models, which clusters Web access logs for operational systems based on access patterns. In this case, the knowledge helps to discover unexpected access paths corresponding to ill-formed Web site design. Third, we have successfully identified undiscovered research issues such as dynamic page recommendation when we have attempted to mine Web usage logs for operational systems.

On the Effectiveness of Web Usage Mining for Page Recommendation and Restructuring

The more pages Web sites consist of, the more difficult the users find it to rapidly reach their own target pages. Ill-structured design of Web sites also prevents the users from rapidly accessing the target pages. In this paper, we describe two complementary approaches to Web usage mining as a key solution to these issues. First, we describe an adaptable recommendation system called the system L-R, which constructs user models by classifying the Web access logs and by extracting access patterns based on the transition probability of page accesses and recommends the relevant pages to the users based both on the user models and the Web contents. We have evaluated the prototype system and have obtained the positive effects. Second, we describe another approach to constructing user models, which clusters Web access logs based on access patterns. We also have found that the user models help to discover unexpected access paths corresponding to ill-formed Web site design.

On mining XML structures based on statistics

We propose an approach to dynamically generate database schemas for well-formed XML data. Our approach controls the number of tables to be divided based on statistics of XML so that the total cost of processing queries is reduced. We devise schemas appropriate for complex data such as text formatting and child elements with the small maximum number of occurrences in order to reduce the number of tables. To this end, we define three functions NULL expectation, Large Leaf Fields, and Large Child Fields for controlling the tables to be divided. We evaluated typical XML queries over the generated schemas and normalized schemas and measured and compared both of the costs. Through this, we successfully validated our approach.

A decentralized XML database approach to electronic commerce

Decentralized XML databases are often used in electronic commerce (EC) business models such as e-brokers on the Web. To flexibly model such applications, we need a modeling language for EC business processes. To this end, we have adopted a query language approach and have designed a query language, called XBML, for decentralized XML databases used in EC businesses. The authors explain and validate the functionality of XBML by specifying e-broker business models and describe the implementation of the XBML server, focusing on the distributed query processing.

Querying Web distributed databases for XML-based e-businesses: requirement analysis, design, and implementation

Electronic commerce (EC) business models like e-brokers on the Web use XML databases such as product and customer data. To flexibly model such applications, we need a modeling language for EC businesses, that is, business processes. To this end, we have adopted a query language approach to modeling and have designed a query language for distributed XML databases called XBML suitable for EC businesses. We discuss the requirements for an XML query language for supporting EC business models and explain the functionality of XBML by specifying e-broker models and describe the implementation of the XBML server.

Building a Navigation Structure from a Fuzzy Relationship for Image Retrieval

Going one step further feedback querying in integrating user into retrieval process, navigation is the more recent approach to find images in a large image collection using contentbased information. However, while properties extracted from images are usually fuzzy data, most of the time a navigation structure will deal with binary links from an image (or a group of images) to another. A trivial solution to get a binary relationship from fuzzy data is to apply a threshold, but this solution not only leads to a loss of information but fails to distinguish noise from interesting elements. In this paper, we propose two techniques to eliminate isolated elements and lead to a structure made of more compact subparts. The first one is based on a variable threshold depending on the number of neighbours. The second one, specific to Galois’ lattice, is based on taking into account the existing navigation structure for binarisation of descriptions. Experiments showed that it improves the resulting structure by reducing the number of nodes without loosing information on image description, thus improving user experience.