Title of Project:

Strengthening the Social Fabric of Singapore: Perspectives From Intergroup Contact Theory

Date:

Aug. 2021 – Present

Funding:

Co-Principal Investigator; Funding from DSO National Laboratories

Co-investigators:

Dr. Nuri Kim (Principal Investigator), Dr. Benjamin Li Junting, Dr. Benjamin Detenber, and Dr. Poong Oh

Description:

This project aims to examine societal faultlines in Singapore and test theory-based strategies to address them. First, we will systematically assess the emerging issues of immigration and LGBT using cutting-edge social scientific approaches to gain a better understanding of the scope and extent of these new societal faultlines. Specifically, public perceptions, sentiments, and attitudes towards marginalized groups expressed in anonymous online spaces will be analyzed. Second, using the theoretical framework of intergroup contact theory (Allport, 1954; Pettigrew, 1998) and the learning anxiety model (Paolini, Harris, & Griffin, 2016), we test several visual and narrative strategies that can help overcome anxieties, improve mutual understanding, and contribute to social cohesion. Finally, we will develop and test VR based applications towards these ends. In doing so, the proposed research aims to contribute to strengthening the social fabric of Singapore by guarding against societal divides.

Research Teams:

Title of Project:

ROSE (Reputable Opinion Social Environment) via Connectors, Mavens and Salesmen

Date:

Feb. 2012 – Jan. 2015

Funding:

Co-Principal Investigator; S$461,850 including one Ph.D. student, Academic Research Fund (AcRF) Tier 2

Co-investigators:

Dr. Chang Kui-Yu (PI), Dr. Zhang Jie, Dr. Huang Weihong, and Dr. Chan Alvin

Description:

In the project, we are building an opinion sharing social network called ROSE (Reputable Opinion Social Environment), which is designed with the sole goal of sharing and mining of trusted reviews among friends within a social network. ROSE allows users to search and compare trusted opinions on anything, including people, products, events, public policies, brands, services, etc. For every product, two rating summaries are computed, 1) average ratings from your friends (in your social network) and 2) average ratings from everyone. For example, a sophisticated science fiction movie like “2001: A Space Odyssey” might be rated higher among your friends who are likely to share the same interest, compared to a lower generic rating from the populace. With ROSE, users no longer have to read through hundreds of peer reviews to make their shopping decisions. Instead, he simply uses ROSE to search for the products via his mobile phone or PC and shall receive a neatly summarized ROSE report card that shows friend and generic sentiment ratings for top features of the product. Users can also compare multiple products side-by-side.

Title of Project:

Digital Intelligence (Phase 2)

Date:

Jul. 2009 – Present

Funding:

Principal Investigator; WKWSCI Academic Research Cluster Fund

Co-investigators:

Dr. Foo Shou Boon, Schubert, Dr. Christopher Khoo, Dr. Theng Yin Leng, and Dr. Chang Yun-Ke

Description:

The main goal of Digital Intelligence research cluster is to develop various text processing and text mining techniques and implement useful web-based intelligent systems. In current Digital Intelligence project, we are focusing on the opinion mining and analysis of social media content. In the beginning, sentiment analysis algorithms were investigated for a non-medical domain, i.e. movie reviews. Then we have switched from the general domain to health care informatics area, and preliminary studies have been done mainly for investigating characteristics of the user reviews on health-related websites. This project follows the current research direction, and focuses on development of an effective method for sentiment analysis and summarization of social media content in health and medical fields. As a target domain, we are using the user reviews of drugs and health supplements (e.g., tamiflu and glucosamine).

Title of Project:

ROSE (Reputable Opinion Social Environment) via Connectors, Mavens and Salesmen

Date:

Feb. 2012 – Jan. 2015

Funding:

Co-Principal Investigator; S$461,850 including one Ph.D. student, Academic Research Fund (AcRF) Tier 2

Co-investigators:

Dr. Chang Kui-Yu (PI), Dr. Zhang Jie, Dr. Huang Weihong, and Dr. Chan Alvin

Description:

In the project, we are building an opinion sharing social network called ROSE (Reputable Opinion Social Environment), which is designed with the sole goal of sharing and mining of trusted reviews among friends within a social network. ROSE allows users to search and compare trusted opinions on anything, including people, products, events, public policies, brands, services, etc. For every product, two rating summaries are computed, 1) average ratings from your friends (in your social network) and 2) average ratings from everyone. For example, a sophisticated science fiction movie like “2001: A Space Odyssey” might be rated higher among your friends who are likely to share the same interest, compared to a lower generic rating from the populace. With ROSE, users no longer have to read through hundreds of peer reviews to make their shopping decisions. Instead, he simply uses ROSE to search for the products via his mobile phone or PC and shall receive a neatly summarized ROSE report card that shows friend and generic sentiment ratings for top features of the product. Users can also compare multiple products side-by-side.

Title of Project:

Influence Detection within Blogosphere 

Date:

Apr. 2010 – Present

Funding:

Principal Investigator; NTU/WKWSCI RCC fund

Co-Principal Investigator:

Mr. Tan Kien Weng Luke

Description:

Blogs play a vital role in spreading new ideas and information on the web. Typically, blogs’ content represents the opinion of bloggers about various topics. This project aims to analyze and comprehend the influence and impact that these opinions could wield on the general blog readers. The project leverages on techniques using content analysis (sentiment analysis), graph theory (blog features), and community detection as possible tools and methods to meet its objective of detecting the presence and propagation of friendly/antagonistic sentiments within blogosphere.

Title of Project:

Mining of Product Comparisons from On-Line Review Documents 

Date:

Apr. 2010 – Present

Funding:

Principal Investigator; NTU/WKWSCI RCC fund

Co-investigators:

Dr. Christopher Khoo

Description:

The main aim of this project is developing a novel algorithm for automatically mining product comparison information. Many researchers have worked on opinion mining and sentiment analysis of review documents. Most of them are focusing on the analysis of one product or subject, such as finding liked/disliked features of a product. Only a few researchers are working on the mining of product comparisons which is a more challenging area. A closely related work by Jindal and Liu used sequential pattern mining with part-of-speech tagging to extract comparison information from comparative sentences. We plan to use various deeper linguistic features, such as functional dependencies of words and semantic tags, to improve relationship extraction patterns. The mined information will have a tabular format where it provides pros and cons of a product in comparison to a specific competing product.

Title of Project:

Sentiment Summarization of Multiple Genre Review Documents

Date:

Mar. 2009 – Feb. 2011

Funding:

Principal Investigator; Academic Research Fund (AcRF) Tier 1

Co-investigators:

Dr. Christopher Khoo and Dr. Chan Syin

Description:

The objective of this project was to develop an effective method for sentiment summarization of multiple review documents from various genres and to implement the method in a sentiment entity search engine. With the explosion of Web 2.0 platforms, enormous amounts of opinionated documents are posted on the Web, including expert reviews, user reviews, blog postings, and discussion board postings which express opinions about movies, electronic products, and social issues. The project focused on analyzing movie reviews, which present interesting challenges as they are written at various levels of complexity by different user groups. The main contribution of this project was the development of a novel method for sentiment summarization using genre-specific sentiment analysis, and its application to implement a sentiment entity search engine.

Title of Project:

Genre and Sentiment Classification of Web Search Results

Date:

Apr. 2008 – Nov. 2009

Funding:

Principal Investigator; NTU/SCI RCC fund (RCC14/2007/SCI, M52069072)

Co-investigators:

Dr. Christopher Khoo

Description:

In this project, only snippets (summary information that includes the URL, title, and summary text) and not full text documents are used in the classifications since full text documents would need more processing time. Determining whether a snippet is a review or non-review document (genre classification) and a positive or negative review document (sentiment classification) is a challenging task since the snippet usually does not contain many useful features for identifying review documents and analyzing sentiments described in documents. The project applies a common machine learning technique, SVM (Support Vector Machine), and heuristic approaches to investigate how effectively the snippets can be used for genre and sentiment classification.

Title of Project:

Automatic Identification of News Frames using Machine-Learning Techniques

Date:

Feb. 2005 – Feb. 2007

Funding:

Principal Investigator; NTU/SCI RCC fund

Co-investigators:

Dr. Christopher Khoo and Dr. Randolph Kluver

Description:

This project develops techniques and a software tool for automatic news frames analysis¾automatically analyzing news articles and categorizing them into one of several pre-defined news frames. News framing analysis is a kind of content analysis of news articles to identify how the news is framed, including the perspective in which the events are reported, how information is selected and organized in the news article, and how the information is expressed. News frames analysis is intellectual work usually performed by human analyzers. The tremendous number of news articles to be analyzed makes manual news frame categorization a difficult and tedious task. This project thus seeks to develop a method for automatic news frame categorization using machine-learning and text mining techniques.

Title of Project:

Information Extraction and Ontology Learning Tools

Date:

Jul. 2003 – Present 

Co-investigators:

Dr. Christopher Khoo

Description:

This project develops ontology learning tools (such as a semantic relation extraction tool for medical documents) using natural language processing, information extraction, and machine learning technologies. The project focuses on the following areas: 1) Usage of an existing medical ontology (such as Unified Medical Language System) for medical domain ontology building, 2) Automatic pattern learning for mining semantic relationships, 3) Semantic relation (e.g., treatment relation) mining from medical abstract documents, and 4) Ontology enrichment and ontology-based digital library application development.

Title of Project:

caT (context-aware Trellis) System

Date:

Jun. 2002 – Dec. 2007

Co-investigators:

Dr. Richard Furuta, Computer Science Department, Texas A&M University, USA.

Description:

caT is a context-aware hypertext model with associated tools, which supports flexible user (or agent) adaptation to changes in environmental information, such as location, time, and bandwidth/cost; caT was developed as my Ph.D. dissertation work. The project focused on the following areas: 1) Multiple browser development (text, images, audio, and video browsers), 2) Large hypertext development support (higher-level authoring specification), and 3) Parallel browsing via multiple devices, such as desktop computers, PDAs, and mobile phones (customizing the content and media type for achieving optimal effect on the target device).