Title of Project: MobiTOP: A System for the Mobile Tagging of Objects and People.

Investigators: A/P Dion Goh, A/P Theng Yin Leng, A/P Lim Ee Peng (SCE), Ast/P Sun Aixin (SCE), A/P Kalyani Chatterjea (NIE), Ast/P Chang Chew Hung (NIE)

Funding: A*STAR

Description: An A*STAR funded project to develop techniques for the creation, management, analysis and discovery of mobile tags, which are media-rich information applied to real-world objects and people. Research areas include user profiling, tag modeling and recommendation, and user interface design. These deliverables will culminate in the implementation of a mobile tagging system known as MobiTOP (Mobile Tagging of Objects and People). Working with pedagogy experts, the system will be deployed and tested in the context of geography education. The project draws upon earlier G-Portal work on geospatial data management and visualization.

Investigators: A/P Theng Yin Leng

To reduce the use of extensive and time-consuming real users testing ubiquitous learning systems, a tool is being developed to automatically generate executable cognitive user models to simulate a real user’s behaviour, as a cost-effective means to rapidly iterate and test system design and detect usability problems in web-based systems. Executable cognitive user models are software agents that simulate real end-users’ behaviour, as well as predict end-users’ performance. The objectives of the project are: 1. To investigate the potential of embedding theories and models of human cognition and artificial intelligence in a tool for constructing executable cognitive user models; 2. To specify the requirements of such a tool for an effective and practical evaluation of web-based systems; 3. To determine how executable cognitive user models can be investigated using software agent technologies throughout the design process; and 4. To investigate how executable cognitive user models can be effectively combined with user testing to achieve the best results.

Investigators: A/P Theng Yin Leng & A/P Dion Goh

Funding: NTU AcRF

Description: Institutions are spending millions of dollars implementing digital libraries (DLs) and Web portals. However, many studies have found the usability and effectiveness of current DLs and portals to be poor. Although there has been some research conducted over the last few years in understanding user needs of text-based and geospatial DLs, there is little work done in helping to make the usability evaluation process of DLs less cumbersome and tedious. Better tools and techniques are needed to help DL designers evaluate their systems in ways that will improve usability to enhance users' experience of DL collections and products. This project investigates usability engineering techniques, a combination of qualitative and quantitative techniques, applicable not only for text-based DLs but also for geospatial DLs, on the Web as well as the mobile environments. DLs of universities, public libraries and national libraries have large user populations, in tens and hundreds of thousands of users. Improvements in DL design can have a major organisational, national and international impact.

Collaborators: Recognising its importance, this proposal has the support of the NTU library and the National Library Board (NLB). Two research centres at NTU, Centre for Human Factors and Ergonomics (CHFE, MPE) and Centre for Advanced Computer Information Systems (CAIS, SCE), and the University of Waikato (New Zealand), are internal and external collaborators working with the project team to exploit the potential of applying this research to the mobile environment, which is fast becoming the popular platform for systems delivering "on-demand" use.

Investigators: A/P Dion Goh & Prof Schubert Foo

Funding: WSCI and NTU Library

The ACRC plans to transform itself into an important regional hub and one-stop center housing quality resources in the specialized areas of Media, Communication and Information. The purpose of the project is develop a digital library for the ACRC to host its electronic collection that includes grey literature, published literature in media, communication and information, and to use the digital library as a platform for conducting research in areas such as information retrieval, information extraction, information organization, data/text mining, collaborative systems, knowledge sharing, etc. The digital library will also provide a platform to support knowledge sharing and publishing by capturing, preserving and communicating the intellectual output of SCI’s faculty staff and researchers. Such a DL system can be further exploited to distribute SCI’s digital works over the Web through a sophisticated search and retrieval system. Availability and easy accessibility of ACRC sources for local, regional and international users would certainly enhance the image of NTU in general, and SCI in particular.

Investigators: A/P Chris Khoo, A/P Dion Goh & A/P Chan Syin (School of Computer Engineering)

Funding: NTU AcRF

A project to develop a prototype intelligent search agent that performs information extraction and synthesis on the Web. Most Web search engines and intelligent search agents merely identify potentially relevant documents on the Web without actually extracting the relevant information from the text of the documents. Information extraction systems developed so far require large training sets, are usable only by experts and take a long time to train. The study seeks to develop an intelligent information extraction system that can be trained by ordinary users using a small number of examples to extract relevant information from multiple Web sites and integrate the information into a multi-document summary to aid in knowledge discovery and knowledge acquisition.

Investigators: A/P Chris Khoo, A/P Dion Goh & Dr Paul Wu

Funding: NTU/SCI RCC

The objective of this study is to develop a method for automatic summarization of sets of sociology abstracts that might be retrieved by a digital library system or search engine in response to a user’s query. The purpose of the multi-document summarizer is to present an overall summary of the set of documents, highlighting the important concepts and relations found in them. The method includes an automatic analysis of the discourse structure of sociology abstracts, both at the macro-level (between sentences and sections) and the micro-level (within sentences). The automatic summarizer focuses on the extraction of variables and semantic relationships between variables expressed in the text, and the integration of the extracted information into a coherent summary.

Investigators: A/P Chris Khoo, Ast/P Na Jin Cheon & A/P Chan Syin (School of Computer Engineering)

Funding:

This project seeks to extend automatic information extraction technology and apply it to the medical domain to extract disease-treatment information from medical abstracts to support evidence-based medicine and knowledge discovery. Current information extraction systems make use of linguistic patterns and pattern matching to identify the pieces of information to extract from unstructured text. The extracton patterns are often constructed automatically by applying a supervised learning technique on a set of manually annotated training text. This project seeks to develop a technique to construct the information extraction patterns without manual annotation of text by performing text mining, automatic text annotation and pseudo-supervised learning. The objectives of the project are:

1. To develop an effective method to mine information extraction patterns in a medical database
2. To develop a method to construct information extraction patterns using pseudo-supervised learning and automated annotation of training text
3. To develop a disease-treatment ontology to model and represent treatment information found in medical abstracts, and to summarize the information to support evidence-based medicine.

Investigators: A/P Chris Khoo & Ast/P Na Jin Cheon

The objective of the project is to develop techniques for automated or computer-assisted sentiment analysis of various genres of text. Sentiment refers to a person’s feeling, emotion or attitude toward a subject, and can cover a variety of emotional dispositions (e.g. anger, admiration, dislike, eagerness, etc.). The appraisal theory (Rothery, 1997; Martin, 1995), which is based on the principles of Systemic Functional Linguistics, is adopted as a framework for the study for its clear explication of how sentiment is expressed in language. It divides appraisal into Attitude, Engagement and Graduation, with Attitude further divided into Affect (emotion), Judgment (ethical/social evaluation) and Appreciation (aesthetic assessment). Current work is focused on:

• automatic categorization of product reviews into positive (favorable/recommended) versus negative (unfavorable/not recommended) sentiment
• development of a sentiment meta-search engine to identify documents and document snippets reporting product reviews and categorizing them into positive and negative reviews
• automatic sentiment analysis of polical news articles using a framework based on the appraisal theory.

Investigators: Dr Paul Wu

Funding: NTU/SCI RCC

The purpose of the project is to develop a Web-based software tool and graphical interface to enable researchers to mark-up and annotate text, encode the annotation in an XML format, store the annotation for further processing, and display the annotation in a number of visual formats. The text annotation tool will be designed to be general and powerful enough to handle most types of content analysis and linguistic analysis. The tool will handle several independent layers of annotation, hierarchical annotation (where primitive units are grouped to form more complex units), and overlapping annotations. Such an annotation tool will be useful for many types of research – content analysis, linguistic analysis, text analysis, creating training documents for text mining and information extraction, etc. A powerful text annotation tool, grounded on a good representation formalism, is needed because a deeper level of content analysis involves a deeper level of linguistic coding. The validation of content analysis results also requires evidence presented in linguistic coding.

Investigators: Dr Paul Wu

Funding: National Library Board

The Internet has increased the proliferation of online publication and community worldwide. Due to the fragility of digital medium, new approach needs to be developed for preservation for future generations; a task imperative in capturing a record of contemporary digital culture and heritage. This project develops a digital repository and Web annotation and cataloguing system for Web archives. By applying intuitive mechanism, evidence of the subject matter and contextual information will be captured as metadata. The metadata further serves as evidence to monitor substantial changes of websites. In sum, the objectives of the project are two fold:
• Evidence-based cataloging: Allow users to effectively catalog websites collection for archival and preservation purposes, reducing the turn-around time, producing verifiable catalogue data/metadata
and thus, increasing the quality of the catalogue.
• Dynamic web content monitoring: Minimize the manual efforts required to maintain the catalogue data/metadata of the web archives through automatically verifying and monitoring the dynamic changes of websites, filtering away unnecessary attention paid to scrupulous changes and alerting only substantive ones that need to be attended to.