Dongyuan has been with School of Computing, National University of Singapore, where she is currently a Research Assistant in Prof. Min Yen Kan's group (WING). She recieved her Ph.D. in July 2012 from Chinese Academy of Sciences, Institue of Automation under the supervision of Prof. Dai Ruwei.

Her general research interests include social network mining, social media analysis and information retrieval, specially in user interest modeling for personalized applications. She has published several conference and journal papers in these fields, including Decision Support System (DSS), WWW, etc.

Here is her updated C.V. (Sep. 2012)


Current Research

Currently, I am interested in user interest modeling for personalized applications, by aggregating both user's contributed content and his social information.

Past Research in PH.D.

  • Hierarchical Object Organization

    For semantic relationship mining, this work is along an alternative research line of organizing web objects into hierarchical semantic categories. An old proverb goes that “I will know what I desire when I saw it.” There is a big chance when users only come with an expectation on what they need but have no clear descriptions. In this case, instead of responding to exact queries, it is more reasonable to facilitate users with a quick overview and enable effective browsing features. Hierarchical organized structure (i.e., tree) provides such a tool as guiding users to the desired objects from coarse level to refined level. Hierarchical structure is very common in web objects. Take Flickr group for example, some groups involve with broader topics (e.g., ‘bird’), while some groups involve with more specific topics (e.g., ‘eagle’). I have worked on organizing Flickr groups into a semantic hierarchy. The premise behind this work is to embed the broader topics at the higher level of the hierarchy and more specific topics at the lower level.

    Here is a copy of the published paper COPY.

  • Sentiment Mining in News

    This research aims at ranking news according to its mood information (e.g., sad, happy, angry, etc). Most of the related studies focus on analyzing the mood information from the writers’ view. While as aforementioned, mood voting enabled by Web 2.0 sites provides an opportunity to organize the objects from the perspective of readers’ mood. This enables users to browse objects depending on their current mood. Since it usually takes a long term to collect votes from readers, this work aims to predict the readers’ mood using a supervised method.

    Here is a copy of the published paper COPY.

  • Personalzied Search

    This work aims at incorporating multiple resources to construct user interest profile. The relationship of user-tag and user-group are encoded as two correlated bipartite graphs, and then spectral graph partitioning is performed to extract the latent interest dimensions. Through this algorithm, user interest can be represented by a vector in the latent interest space, and each dimension of the interest space can be interpreted by tags or groups. This model well captures the diversity nature of user interest as well as topic-sensitive property of social links. The constructed user interest model is applied to improve persoanlized sesearch application.

    Here is a copy of the published paper COPY.

  • Prestigious Member Identificationa and Prediction

    Since users inevitably interact with each other in a social network scenario, to understand the social behaviors is crucial for user modeling. Prestigious user identification and prediction is a typical problem in social network analysis (SNA). It involves with finding the current and future most influential users within a community, which is very important to investigate behavior patterns and social trends. This research work proposed a framework of graph-based action network for indetifying and predicting prestigious members in Flickr groups.

    Here is a copy of the published paper COPY.



  1. Dongyuan Lu, Qiudan Li and Stephen Liao (2012). A graph-based action network framework to identify prestigious members through member’s prestige evolution in Flickr groups, in Decision Support System (DSS), Volume 53, Issue 1, 44-54. DOI=10.1016/j.dss.2011.12.003. [Local Copy]
  2. Jitao Sang, Changsheng Xu and Dongyuan Lu (2012). Learn to personalized image search from the photo sharing websites, in IEEE Transactions on Multimedia (TMM). Volume: 14, Issue 4, 963-974. DOI=10.1109/TMM.2011.2181344
  3. Dongyuan Lu and Qiudan Li (2011). A novel news recommendation method by aggregating reader’s mood, in Journal of Chinese Information Processing. Volume 25, Issue 3, 79-85 (in Chinese) [Local Copy]


  1. Dongyuan Lu and Qiudan Li (2011). Personalized search on Flickr based on searcher's preference prediction. In Proceedings of the 20th international conference companion on World wide web (WWW '11). ACM, New York, NY, USA, 81-82. DOI=10.1145/1963192.1963234 [Local Copy]
  2. Dongyuan Lu and Qiudan Li (2010). Exploiting Semantic Hierarchies for Flickr Group. In Proceedings of 6th International Conference, AMT 2010. Toronto, Canada, 74-85. DOI=10.1007/978-3-642-15470-6_9 [Local Copy]


Publication List Assistant

Help you manage your publication list.


Dongyuan Lu (Dr)

Research Assistant

  • Email:
  • Address: AS6 05-21, School of Computing, National University of Singapore 13 Computing Drive Singapore 117417