Biography

kenneth Kenneth Wai-Ting Leung received the B.Sc. degree in Computer Science from the University of British Columbia, Canada, in 2002. He then obtained his M.Sc. and Ph.D. degrees in Computer Science and Engineering from the Hong Kong University of Science and Technology in 2004 and 2010, respectively. He is currently a Visiting Assistant Professor in the Department of Computer Science and Engineering at the Hong Kong University of Science and Technology. His research interests include information retrieval and mobile computing. More specifically, he is interested in topics involving clickthrough clustering, personalized web search, mobile web search, and collaborative web search. The major aim of his research is to improve a search engine's retrieval effectiveness so that relevant information can be easily discovered by the users. His CV is available here.

Publications

  • Leung, K.W.-T., Lee, D.L., and Lee, W.-C., PMSE: A Personalized Mobile Search Engine, IEEE Transactions on Knowledge and Data Engineering (TKDE). (To Appear)
  • Vosecky, J., Leung, K.W.-T., Ng, W., Searching for Quality Microblog Posts: Filtering and Ranking based on Content Analysis and Implicit Links, Proc. of DASFAA Conference, Busan, South Korea. (To Appear, Acceptance Rate 27.6%)
  • Leung, K.W.-T., Lee, D.L., Ng, W., and Fung, H.Y., A Framework for Personalizing Web Search with Concept-Based User Profiles, ACM Transactions on Internet Technology (TOIT). (To Appear)
  • Jiang, D., Leung, K.W.-T., and Ng, W., Context-Aware Search Personalization with Concept Preference, Proc. of CIKM Conference, Glasgow, Scotland, 2011. (Acceptance Rate 15%)
  • Leung, K.W.-T., Lee, D.L., and Lee, W.-C., CLR: A Collaborative Location Recommendation Framework based on Co-Clustering, Proc. of SIGIR Conference, Beijing, China, 2011. (Acceptance Rate 19.8%)
  • Leung, K.W.-T., Fung, H.Y., and Lee, D.L., Applications of Concept Relation Network to Web Search, International Workshop on Linked Web Data Management, Uppsala, Sweden, 2011.
  • Leung, K.W.-T., Fung, H.Y., and Lee, D.L., Constructing Concept Relation Network and its Application to Personalized Web Search, Proc. of EDBT Conference, Uppsala, Sweden, 2011.
  • Leung, K.W.-T., and Lee, D.L., Dynamic Agglomerative-Divisive Clustering of Clickthrough Data for Collaborative Web Search, Proc. of DASFAA Conference, Tsukuba, Japan, 2010, 635-642. (Acceptance Rate 30%)
  • Leung, K.W.-T., Lee, D.L., and Lee, W.-C., Personalized Web Search with Location Preferences, Proc. of IEEE ICDE Conference, Long Beach, California, USA, 2010, 701-712. (Acceptance Rate 12.5%)
  • Leung, K.W.-T., and Lee, D.L., Deriving Concept-based User Profiles from Search Engine Logs, IEEE Transactions on Knowledge and Data Engineering (TKDE), Vol. 22, No. 7, Jul 2010, 969-982.
  • Leung, K.W.-T., Ng, W., and Lee, D.L., Personalized Concept-Based Clustering of Search Engine Queries, IEEE Transactions on Knowledge and Data Engineering (TKDE), Vol. 20, No. 11, Nov 2008, 1505-1518.
  • Lee, D.L., Leung, K.W.-T., Ng, Lee, W.-C., Chan, T., Chan, T.M., and Iu, K.H., Performance of Preference Mining Methods on Search Engines, 1st International Conference on Ubiquitous Information Management and Communication (ICUIMC), Seoul, Korea, 2007. (Invited Paper)
  • Leung, K.W.-T., Lee, D.L., Ng, W., Lee, W.-C., Chung, T.K., Ma, P.Y., Tse, C.C., and Wong, C.Y., Discovering User Communities from Clickthough Data, 1st International Conference on Ubiquitous Information Management and Communication (ICUIMC), Seoul, Korea, 2007. (Invited Paper)
  • Chan, V.W., Leung, K.W.-T., and Lee, D.L., Clustering Search Engine Query Log Containing Noisy Clickthroughs, Proc. of SAINT Conference, Tokyo, Japan, 2004, 305-308.

Professional Activities

  • Conference Reviewer for VLDB 2007, DEXA 2008, FoIKS 2008, WAIM 2008, WISE 2008, ACM GIS 2009, ACM SAC 2009, CIKM 2009, FoIKS 2009, MDM 2009, WAIM 2009, FoIKS 2010, ICDE 2010, DEXA 2010, KES 2010, MDM 2010, PAKDD 2010, ACM SAC 2011, ECIR 2012
  • Journal Reviewer for ACM TOIS, ACM TSIT, IEEE ICSI, IEEE SMCA, IEEE TKDE
  • Member of the Golden Key International Honour Society

Awards

  • Oversea Research Award (2010, award scheme that offers selected students to work in an oversea institution).
  • Research Travel Grant (2010, grant award to support the cost of my travel to IEEE ICDE 2010).
  • Postgraduate Studentship (2006-Present, studentship are allocated to full-time research postgraduate students).
  • First Class Standing (2002, Overall GPA > 85% in the University of British Columbia).
  • Undergraduate Scholar Program Scholarship (2001, an academic average of at least 85.00% on best 27 credits)
  • UBC Dean's Honour List (2000-2001, an academic average of 85% or higher)
  • J Fred Muir Memorial Scholarship in Science (2000, offered to students in the Faculty of Science on the recommendation of the Faculty)
  • British Columbia Government Scholarship (1998, offered to students with outstanding performance on the provincial examinations)

Teaching

  • 2011-2012 Spring - Shadowing Program with Fukien Secondary School, 1 week collaborative program with Fukien Secondary School to involve and supervise secondary school students in HKUST research and study
  • 2011-2012 Spring - COMP4021 Internet Computing
  • 2011-2012 Fall - COMP6311 Topics in DB: Mobile and Location-Based Search (Co-taught with Prof. Dik Lun LEE)
  • 2011-2012 Fall - COMP2711 Discrete Mathematical Tools for Computer Science
  • 2011-2012 Fall - COMP4021 Internet Computing
  • 2011-2012 Fall - COMP4431 Multimedia Computing
  • 2010-2011 Spring - COMP343 Multimedia Computing
  • 2010-2011 Spring - COMP170 Discrete Mathematical Tools for Computer Science
  • 2009-2010 Spring - COMP630 Mobile Search Engine
  • 2006-2007 Spring - COMP630 Internet and Mobile Information Retrieval
  • 2006-2007 Spring - CSIT531 Information Retrieval on Internet and Mobile Networks
  • 2006-2007 Spring - COMP336 Search Engines for Web and Enterprise Data
  • 2006-2007 Fall - COMP104H/171H Programming Fundamentals and Methodologies/Data Structures and Algorithms (Honored Track)
  • 2001-2002 Fall - CPSC152 Principles of Software Development

Implemented Prototypes

Research Interests

My research interests lie in the area of information retrieval and mobile computing. Specifically, I am interested in topics involving clickthrough clustering, personalized web search, mobile web search, and collaborative web search. The major aim of my research is to improve a search engine's retrieval effectiveness so that relevant information can be easily discovered by the users.

  1. Clickthrough Clustering
    During my master degree, I focused on studying query clustering techniques to discover similar queries on a search engine, such that query suggestions can be provided to users to formulate more effective queries. I had collaborated with my supervisor, Prof. Dik Lun Lee, on a query clustering project to implement the agglomerative algorithm proposed by Beeferman and Berger. After the implementation, I determined a weakness of the method with noisy user clickthroughs. To tackle the problem, I proposed a new noise-tolerant similarity function for the algorithm. I also observed that most major commercial search engines provide query suggestions to help users formulate more effective queries. When a user submits a query, a list of terms that are semantically related to the submitted query is provided to help the user identify terms that he/she really wants, hence improving the retrieval effectiveness. Yahoo's "Also Try" and Google's "Searches related to" features provide related queries for narrowing search, while Ask Jeeves suggests both more specific and more general queries to the user. Unfortunately, these systems provide the same suggestions to the same query without considering users' specific interests. Thus, I proposed a method that provides personalized query suggestions based on a personalized concept-based clustering technique. The method not only identify the objects that users are interested in (i.e. positive preferences), but also the objects that users dislike (i.e. negative preferences). In contrast to existing methods that provide the same suggestions to all users, my approach used clickthrough data to estimate user's conceptual preferences and then personalized query suggestions were provided for each individual user according to his/her conceptual needs.

  2. Personalized Web Search
    Personalized web search is an important means to improve the performance of a search engine. Thus in 2008, I proposed a framework that supports mining a user's conceptual preferences from users' clickthrough data resulted from web search. The discovered preferences were utilized to adapt a search engine's ranking function, such that search results relevant to the user's conceptual need will appear higher in the result list. The concept-based user profiles (CUP) were similar to the one proposed in my clustering works. In the framework, an extended set of conceptual preferences was derived for a user based on the concepts extracted from the search results and the clickthrough data. Then, a Concept-based User Profile representing the user profile as a concept ontology tree was generated. Finally, the CUP was input to a Support Vector Machine (SVM) to learn a concept preference vector for adapting a personalized ranking function that re-ranks the search results. In order to achieve more flexible personalization, the framework allowed a user to control the privacy level that specifies how much the CUP information is exposed to the personalized search engine. I studied various parameters, such as conceptual relationships and concept features, arising from CUP that affect the ranking quality. I confirmed that the approach was able to improve significantly the retrieval effectiveness for the user. Further, I showed that the proposed privacy control parameters can adjust the exposed user information more smoothly and maintain better ranking quality than the existing methods.






  3. Mobile Web Search
    In mobile search, the interaction between users and mobile devices are constrained by the small form factors of the mobile devices. To reduce the amount of user's interactions with the search interface, an important requirement for mobile search engine is to be able to understand the users' needs, and deliver highly relevant information to the users. Personalized web search is one way to resolve the problem. By capturing the users' interests in user profiles, a personalized search middleware is able to adapt the search results obtained from general search engines to the users' preferences through personalized reranking of the search results. I recognized the importance of location information in mobile web search and proposed to incorporate the user's location preferences in search personalization. Thus, I proposed an ontology-based, multi-facet (OMF) user profiling strategy to capture both of the users' content and location preferences (i.e., multi-facets) for building a personalized search engine for mobile users.

    According to a leading market research firm, mobile search will account for around $715 million, or almost 15% of a total mobile advertising market worth nearly $4.7 billion by 2011.





  4. Collaborative Web Search
    Mining search engine clickthrough data is very useful for discovering relationships between users, queries, and documents. In this work, I proposed community clickthrough model (CCM) which modeled clickthroughs as a tripartite graph involving users, queries and concepts embodied in the clicked pages. I developed the Dynamic Agglomerative-Divisive Clustering (DADC) algorithm for clustering the tripartite clickthrough graph to identify groups of similar users, queries and concepts to support collaborative web search. Since the clickthrough graph is updated frequently, DADC clusters the graph incrementally, whereas most of the traditional agglomerative methods cluster the whole graph all over again. Interesting user communities were then identified by walking through the clustered CCM tripartite graph. To support collaborative web search, clickthroughs from a user community were gathered to learn a community-based personalized ranking function to rank the search results according to the preferences mined from the user community. Experimental results showed that DADC can discover similar users, queries and documents more effective in terms of precision and recall comparing to existing methods.

Supervised Projects

  1. Summarization of Tweets for Easy Understanding and Interactive Exploration (DL1), 2011-2012
  2. Using Social Knowledge for Collaborative Search (DL2), 2011-2012
  3. Mobile Web Search (DL3-1), 2011-2012
  4. Collaborative and Social Web Search (DL3-2), 2011-2012
  5. SoJetso - A Location-Based Mobile Application (DL1), 2010-2011

    Mobile devices with GPS function become very popular recently. Thus, more and more location-base applications (LBA) are developed for mobile devices, such as iPhone and Android phone. Due to the high inflation in Hong Kong, many people aware of the ways of getting discount, while they are shopping. Therefore, we developed a LBA, namely "SoJetso", on the Google Android platform to facilitate automatic search of nearby discount for people. We adopted several algorithms to predict different suggestions according to the users' shopping preferences.
  6. Collaborative Location Recommender - ProSearch (DL3), 2010-2011

    A location-based service (LBS) is a mobile application that is capable of utilizing the geographical position of the mobile device. A primary function of a location-based service (LBS) is to recommend interesting locations to the users. In this project, we implement a mobile collaborative location recommendation recommendation (CLR), which employs GPS data in location-based service (LBS) to automatically generate collaborative location recommendations.
  7. Community-based Web Search (DL4), 2010-2011

    Many collaborative web search methods are based on user profiling and clustering on user communities. In this project, we propose the Community Clickthrough Model (CCM), which captures users' conceptual preference and thus outperforms other content-ignorant models. The corresponding clustering algorithm on CCM, Community-based Agglomerative Divisive Clustering (CADC) algorithm allows incremental clustering of the clickthrough data. It significantly outperforms existing clustering methods in terms of both speed and accuracy.
  8. Reading a Million Words in One Minute (DL4), 2009-2010

    Nowadays, most people have to browse massive amount of information from web pages and documents. Thus, an effective summarization system is needed in order to help users finding relevant information from the massive amount of data. In this project, we implemented a summarization system that is capable of summarizing a large amount of information into a set of related concepts. A highly customizable interface is employed, such that users can easily roll-up or drill-down to the concepts that are related to their information needs in order to find the relevant information.
  9. Mobile Application with User Interests and Activity Profiling (DL3), 2009-2010
    The goal of this project is to deliver more relevant information to mobile users according to their habits, visited locations and daily activities. In many recommendation systems, user preferences are mainly learned from the content of the users' browsed documents. However, we can gather much more information such as GPS locations, user movements, schelude/calendar,...etc to produce an active user profile. The active profiles were employed to optimize the information to be delivered to the user. In this project, we aimed at developing a personalized mobile assisting application, which consists of an intelligent search application, a smart calendar/scheduler and a highly customizable interface. The application would display useful information according to the information stored in the active user profile.
  10. An Android-based Mobile Search Engine (DL2), 2008-2009

    The project aims at using Google Android SDK to develop a mobile search engine application which can provide personalized search results according to the users' topical and location preferences. The most important component in this project is the user profiling strategy to capture both the users' topical and location interests in ontology-based, multi-facet (OMF) user profiles. After we accurately determined the user preferences in the OMF user profiles, the profiles are used to train a SVM ranker to rank the search results according to the preferences captured in the OMF user profiles. We implemented a mobile search prototype on Google Android platform to validate our proposed strategy.
  11. Community-Based Search Engine (DL1), 2008-2009

    Search Engines has become an essential tool in daily life. Tons of information can be found on the Internet. But how to make the searching become effective also becomes a challenging research subject for search engine developer. In this project, we aims to address the feasibility of utilizing clustering algorithms for grouping user based on their interests and providing them with personalized query suggestions according to their user communities and interests.
  12. An Ontology-based Personalized Search Engine (DL3), 2007-2008

    Ontology and search engine personalization coexist in the research area for a certain while, but seldom is a relationship drawn between them, so are the conceptually equivalent keywords searched by millions of web users daily. This project aims to address the feasibility of utilizing ontology in search engine personalization and research on the optimal method of relating the keywords as concepts in an ontology.
  13. Clustering of Clickthrough Data (DL2), 2006-2007

    Information retrieval on the World Wide Web has become an important issue in recent years. Developing an effective search engine is challenging as the amount of information available on the Internet is huge (sized 10 billion pages in 2006 and doubles yearly). We find that effective keywords in the user queries are needed in order to retrieve the correct information for the users. Thus, we proposed a method to effectively discover similar queries from the clickthrough data, such that users can formulate more effective queries by picking the correct candidates from the similar query list. A prototype search engine has been implemented to automatically link up similar queries, and to provide personalized search results to according to the users' interests.
  14. Personalized Search Engine (DL1), 2006-2007

    Personalized search engine aim at helping users to find relevant information according to their preferences. Different preference mining algorithms have been proposed in order to accurately determine the users' preference. In this project, we studied the preference mining algorithm, SpyNB, presented in "Mining User Preference Using Spy Voting for Search Engine Personalization", by Ng, W., Deng, L., and Lee, D.L. We employed different types of features (e.g. categorization, time on page, click frequency, queryURLOverlap, related image, style tag, and hyperlink relevancy)  on SpyNB, and implemented a full text search engine using Lucene library to evaluate SpyNB again other existing preference mining algorithms.
  15. Location-based Visual Tour (DL4), 2005-2006

    A virtual tour application with location-based service to guide users to their target locations around the HKUST campus. The virtual tour system supports real-time updates such that the user's most up-to-date location would be detected and displayed.
  16. Indoor Floor Plan Modeler and Service Provider (DL3), 2005-2006

    The primary purpose of the project was the development of a digital floor plan server and a floor plan modeler. The whole architecture of the system was built on top of flexible and OGC standard compliance recommendation. The developed system was highly scalable for the representation of various set of floor plan data in diverse situations.
  17. Personalized Search Engine (DL1), 2006-2007

    Applying mobile technology onto daily life is the motivation of the project. Robots should be playing a role of human's assistant which replacing people to perform dangerous and complicated tasks. In this project, we designed a professional robotics system which provides computer vision, hearing, remote control, and autonomous mobility. Moreover, a laptop, which acts as a central processor, is mounted to handle the command and send or receive signal.
  18. Personalized Search Engine (DL1), 2006-2007

    In this project, we proposed a positioning approach based on processing real-time signal strength information available at tetra-base access points to provide overlapping coverage in the area of interest. We made use of techniques that combined experimental results and observed measurements to demonstrate the feasibility of estimating user position with a high degree of accuracy.