Harris Wu

Harvesting social knowledge from folksonomies

Collaborative tagging systems, or folksonomies, have the potential of becoming technological infrastructure to support knowledge management activities in an organization or a society. There are many challenges, however. This paper presents designs that enhance collaborative tagging systems to meet some key challenges: community identification, ontology generation, user and document recommendation. Design prototypes, evaluation methodology and selected preliminary results are presented.

Probabilistic question answering on the Web

Web-based search engines such as Google and NorthernLight return documents that are relevant to a user query, not answers to user questions. We have developed an architecture that augments existing search engines so that they support natural language question answering. The process entails five steps: query modulation, document retrieval, passage extraction, phrase extraction, and answer ranking. In this article, we describe some probabilistic approaches to the last three of these stages. We show how our techniques apply to a number of existing search engines, and we also present results contrasting three different methods for question answering. Our algorithm, probabilistic phrase reranking (PPR), uses proximity and question type features and achieves a total reciprocal document rank of .20 on the TREC8 corpus. Our techniques have been implemented as a Web-accessible system, called NSIR.

The effects of fitness functions on genetic programming-based ranking discovery for Web search

Genetic-based evolutionary learning algorithms, such as genetic algorithms (GAs) and genetic programming (GP), have been applied to information retrieval (IR) since the 1980s. Recently, GP has been applied to a new IR task-discovery of ranking functions for Web search-and has achieved very promising results. However, in our prior research, only one fitness function has been used for GP-based learning. It is unclear how other fitness functions may affect ranking function discovery for Web search, especially since it is well known that choosing a proper fitness function is very important for the effectiveness and efficiency of evolutionary algorithms. In this article, we report our experience in contrasting different fitness function designs on GP-based learning using a very large Web corpus. Our results indicate that the design of fitness functions is instrumental in performance improvement. We also give recommendations on the design of fitness functions for genetic-based information retrieval experiments.

Probabilistic question answering on the web

Web-based search engines such as Google and NorthernLight return documents that are relevant to a user query, not answers to user questions. We have developed an architecture that augments existing search engines so that they support natural language question answering. The process entails five steps: query modulation, document retrieval, passage extraction, phrase extraction, and answer ranking. In this paper we describe some probabilistic approaches to the last three of these stages. We show how our techniques apply to a number of existing search engines and we also present results contrasting three different methods for question answering. Our algorithm, probabilistic phrase reranking (PPR) using proximity and question type features achieves a total reciprocal document rank of .20 on the TREC 8 corpus. Our techniques have been implemented as a Web-accessible system, called NSIR.

Research on e-commerce transaction networks using multi-agent modelling and open application programming interface

We provide a formal definition of an e-commerce transaction network. Agent-based modelling is used to simulate e-commerce transaction networks. For real-world analysis, we studied the open application programming interfaces (APIs) from eBay and Taobao e-commerce websites and captured real transaction data. Pajek is used to visualise the agent relationships in the transaction network. We derived one-mode networks from the transaction network and analysed them using degree and betweenness centrality. Integrating multi-agent modelling, open APIs and social network analysis, we propose a new way to study large-scale e-commerce systems.

Mining web navigations for intelligence

The Internet is one of the fastest growing areas of intelligence gathering. We present a statistical approach, called principal clusters analysis, for analyzing millions of user navigations on the Web. This technique identifies prominent navigation clusters on different topics. Furthermore, it can determine information items that are useful starting points to explore a topic, as well as key documents to explore the topic in greater detail. Trends can be detected by observing navigation prominence over time. We apply this technique on a large popular website. The results show promise in web intelligence mining.

Quality of data standards: framework and illustration using XBRL taxonomy and instances

The primary purpose of data standards is to improve the interoperability of data in an increasingly networked environment. Given the high cost of developing data standards, it is desirable to assess their quality. We develop a set of metrics and a framework for assessing data standard quality. The metrics include completeness, relevancy, and a combined measure. Standard quality can also be indirectly measured by assessing interoperability of data instances. We evaluate the framework on a data standard for financial reporting in United States, the Generally Accepted Accounting Principles (GAAP) Taxonomy encoded in eXtensible Business Reporting Language (XBRL), and the financial statements created using the standard by public companies. The results show that the data standard quality framework is useful and effective. Our analysis also reveals quality issues of the US GAAP XBRL taxonomy and provides useful feedback to taxonomy users. The Securities and Exchange Commission has mandated that all publicly listed companies must submit their filings using XBRL. Our findings are timely and have practical implications that will ultimately help improve the quality of financial data and the efficiency of the data supply chain in a networked business environment.

Supporting process design for e-business via an integrated process repository

Business process design is an integral part of e-business engineering. Given that e-business models usually involve a wide range of business processes across different business functions with complex activities, events, and documents, process design for e-business is a very challenging task. Although various process reference models (PRMs) have been developed to provide guidelines for process design, research on leveraging multiple PRMs to support process design for e-business has been scant. In this paper, we demonstrate that the diverse process design requirements in e-business are best satisfied by utilizing multiple PRMs via a case study. Then, we propose a collaborative approach grounded in knowledge management theory to integrating multiple process reference models to better support process design in e-business. We equip the integrated process repository with a set of novel features based on Web 2.0 technologies to enhance its utility, efficiency, and quality for process design support. A prototype system is developed and user experiments are conducted to evaluate the system.

New Predictive Control Scheme for Networked Control Systems

This paper is concerned with the design of networked control systems with random network-induced delay and data dropout. It presents a new control scheme, which is termed networked predictive control with optimal estimation. Based on Multirate Kalman Filtering, the measured data which are out of sequence or delayed can be used to improve the precision of estimation. The control prediction generator provides a set of future control predictions to make the closed-loop system achieve the desired control performance and the compensator removes the effects of the network transmission with time delay and data dropout. Simulation results are presented to illustrate the effectiveness of the control strategy via comparing with control schemes without any compensation for the network.

Risk identification and conduction model for financial institution IT outsourcing in China

In this paper we study the risks around IT outsourcing in Chinese financial institutions. We build a framework that explains the conduction paths from risk factors to risks, and from risks to resulting losses. Our framework describes relationships among 8 risk factors, 7 risks and 4 kinds of losses. Through case study we find that the main risk factors faced by Chinese financial institutions during IT outsourcing include limited IT literacy, limited choices of contractors, cultural conflict and objective misalignment with the contractors. These factors comprise the main differences between developing and developed countries in their financial institution IT outsourcing practices.