Published on: 08-Jun-2017
Development of an online Forum for Mental Health Related Topics
Author: Tan Jia Lin and Gabriela Maisie Gustaman (Computer Science)
Supervisor: Dr Li Fang
Millions of people suffer from mental health issues. Based on research done in the United States, statistics have shown that 1 in 5 adults experience some form of mental health problems. The number of people that suffer from these problems has risen over the past decade and is set to keep rising.
Despite the large numbers of people suffering from mental illnesses and disorders, studies have shown that more than 50% of them do not seek any professional help. Mental disorders are real problems that can affect the lifestyles of individuals. Most have recognised treatment methods that can alleviate the problems, yet many individuals do not seek regular professional help due to a variety of other issues. There are concerns over costs, confidentiality and social stigma attached to mental disorders, deterring sufferers from seeking professional help and optimal care. As a result, there is a need to address these concerns so that more sufferers can seek help.
The aim of this project is to develop an online community support forum dedicated to mental health issues. The platform will share details about common disorders while allowing those suffering from them to post and share their problems. Strong community support is integral to dealing with mental health issues. Currently, even though several similar websites exist, none have an interactive and clean interface. A user-friendly and easily navigable website would encourage more users to share more content and find information easily. Through the use of the forum, users will be able to seek help from the online community, which, includes other users with similar experiences as well as the forum’s resident psychiatrist.
Additionally, the forum will serve as a medium for safe and anonymous posting, allowing users an alternative method to seek redress. Apart from user created content and discussions, the forum will serve as a resource bank that provides caretakers and families detailed information about the problems to expect and tips to deal with them.
The forum is for everyone to read, find and write content related to mental health issues. The content of the product is user-generated hence, the forum provides a variety of interaction tools such as commenting, following features, notifications as well as liking posts to encourage more user content. The administrators of the system will also have the ability to curate posts to a specific topic.
To help users find the right content they need, the curation would help to filter post according to the most recent issues. Users are also able to search the entire forum by using keywords.
There will also be special users called “Experts” which are certified Psychiatrist to be using the forum to comment on posts and give advice.
The name of the forum is called “Dear Carrie”. This name is to signify that every post is written to the system which is called Carrie as if you are sharing your problems to a friend. The name “Carrie” is used as it contains the word “Care” which is the intention of the forum, to care for people with mental-health related issues.
The figure below shows the initial brain-storming of the logo by playing with different swatches and fonts.
The logo of the forum is a combination of the alliteration of the name of the website. The alphabet “D” and “C” are titled 45 degrees are combined together to form a heart shape to show the “caring” side of the website.
The figure below shows the creation and union of the logo.
SCE YouTube link:
Collusion-resistant Spatial Phenomena Crowdsourcing
Author: Xiang Qikun (Computer Science)
Supervisor: Associate Professor Zhang Jie
Data trustworthiness is a crucial issue in real world crowdsourcing and participatory sensing applications. Without considering this issue, different types of worker misbehavior, especially the challenging collusion attacks, can result in biased and inaccurate estimation and decision making.
In this project, we constructed a novel trust-based mixture of Gaussian processes (GP) model for spatial field regression in participatory sensing to jointly detect worker misbehaviors and reconstruct the spatial field. We developed a Markov chain Monte Carlo (MCMC)-based inference algorithm to efficiently perform Bayesian inference of the proposed model. The inference algorithm was implemented in MATLAB. We experimentally verified that the proposed model is able to produce more accurate predictions when untrustworthy data is present.
SCE YouTube link:
Author: Freny Nelin (Computer Science)
Supervisor: Dr Owen Noel Newton Fernando
Augmented Reality (AR) and Bluetooth Low Energy (BLE) Beacon technologies are currently the most employed technologies in mobile devices either for educational or entertainment purposes due to their interactive nature. However, most mobile applications utilize only AR or only Bluetooth as the main technology, and therefore are insufficient in bypassing each of the utilized technology’s drawbacks. For image recognition, current mobile applications require good lighting condition and hence perform poorly in environments where varying lighting intensities are used. Likewise, Bluetooth demonstrates weakness as its signal waves are subjected to interference from other radio spectrum operators such as Wi-Fi, thus making it difficult to accurately pinpoint its target in isolation.
Prof. Who is invented with the idea of complementing the drawbacks of image recognition and Bluetooth beacon detection while maintaining the interactivity of the application through seamless integration of the two technologies. This novel hybrid system is implemented to enable the users to retrieve and view school information on professors, posters, and laboratories interactively and easily through an image or a Bluetooth scan.
The scan can be performed from inside or outside NTU. As long as the user has downloaded the application, users who are in NTU are able to scan professors’ office door tags, or Bluetooth signals emitted from a Bluetooth signal emitting beacon attached to professors’ offices while walking past the school office corridor. Users who are outside of NTU are able to scan professors’ profile pictures from school official faculty website or posters from Prof. Who website.
When one of the technologies fails in this application due to unfavourable & undesired situations like dim environments, the application will fall-back to use the other technology automatically or with user’s consent to remain interactive and functional for users. This project was implemented with the aim of improving user experience in retrieving information, without compromising its functionalities.
SCE YouTube link:
Sentiment Analysis using Deep Learning
Author: Raji Cherian Kevin (Computer Science)
Supervisor: Associate Professor Lin Weisi
Emotions and sentiment are the fundamental components of any human interaction and allow us to connect and develop relationships with other humans. This is a fundamental part of the way we interact and live. Integrating the ability to identify emotion and sentiment into the next generation of devices and applications will allow us to have more natural interactions as they will be able to understand you at a deeper level and thereby cater applications and services in accordance with your mood and emotion.
In this project, we have evaluated the effectiveness of different recurrent neural network architectures in predicting sentiment in tweets and predicting emotion in children’s stories. Long Short Term Memory Networks (LSTMs) were purpose built to remember information for long periods of time, and therefore work extremely well in evaluating the entire context of the sentence. By developing several models and running a considerable number of experiments, we have found that Bidirectional LSTMs (BLSTM) offer us the best performance in terms of analysing tweets and predicting emotion in children’s stories. They can accurately model the salient information about the underlying sentiment and emotion in a text input by making use of all available information.
In this project, we have built two systems; one that analyses sentiment on tweets and another that classifies emotion in children’s stories. The fundamental structure of both systems consists of a neural network model that makes use of a Bidirectional LSTM. Recurrent neural networks are capable of modeling sequential information of words. This allows them to be especially effective with narrative text like fairy tales and children’s stories as it can make use of past and future information to predict the class of the emotion.
SCE YouTube link:
Twittener: Listen to what's happening on Twitter
Author: Ng Hong Quan (Computer Science)
Supervisor: Dr Owen Noel Newton Fernando
The use of social media has become increasingly prevalent in the world today. There were 2.3 billion active social media users worldwide in 2016, a 10% increase from the previous year.
The Web significantly facilitates information and service retrieval, particularly for the elderly who may be subject to physical and cognitive function declination which prevents them from carrying out their daily tasks. In addition, the decrease in mobility causes a reduction of social contact that often leads to social isolation and depression.
However, due to their features, characteristics and needs, the elderly often face obstacles in the use of digital technology, such as the Internet. The difficulties faced by the elderly in the use of computers and Internet has been documented, and scientists identified difficulty in reading & understanding texts, recognizing and accessing links, and navigating as frequently cited issues. Apart from the elderly, people with disabilities often face similar difficulties, depending on their mobility & cognitive functioning, when accessing the web. Such difficulties may aggravate loneliness and create isolation for the stricken. Therefore, web accessibility is imperative in building an inclusive society and caring for the well-being of the population. Additionally, aside from supporting the elderly & the disabled, it has benefits for ordinary users who uses the web too.
The social media of choice in this project is Twitter, as it is a popular social networking service used by all kinds of people, especially to keep up with the news, and to engage in social activities such as sharing their thoughts and staying in contact with individuals they are interested in, including celebrities and politicians.
Twittener, the brainchild of this project, provides an alternative way for users to get updates from their feeds, by listening to their feeds instead of reading them. Speech synthesis (Text-to-Speech), identified as an assistive technology, was implemented to convert tweets to speech then delivering the content to users.
Further improvements to the accessibility of the application were made with the development of voice control and input. Speech recognition, together with natural language understanding is used to transcribe user's voice and understand the intentions of the user for the application to perform the actions on behalf of the user. Hands-free interaction is therefore possible with the use of voice instruction to interact directly with the application.
This application could prove to be beneficial for people with disabilities and the elderly to get connected and stay in the loop with what is happening around them. In addition, it can also be useful for people who like a hands-free interaction when multitasking, such as drivers and joggers, by facilitating their interaction with social media, by listening to their Twitter feed & posting to Twitter by voice command. The exciting project that is Twittener holds promise and provides opportunities to increase the uptake of social media in an engaging, and facilitated manner.
SCE YouTube link:
Online shopping with tactile user experience
Chen Ningshuang (center) together with A/P Sourin’s PhD student, Zhang Xingzi have also submitted a research paper on this and accepted by CGI 2017.
Author: Chen Ningshuang (Computer Engineering)
Supervisor: Associate Professor Alexei Sourin
With the increasing popularity of online shopping in the last decade, there has been a trend towards creating a more immersive life-like shopping environment for online shoppers. Haptic technology serves as a potential interaction option for e-commerce companies to adopt. However, the feasibility of incorporating haptic interaction into online shopping has not been proved, nor has the extent of the usefulness of such incorporation.
Thus, this project explores the feasibility and meaningfulness of integrating data-driven haptic feedback into online shopping product browsing experience using a desktop haptic device. In this project, haptic weight, geometry, and material simulation based on real-life measurements are provided as additional means for shoppers to gather product information. A user study was conducted to examine users’ attitude towards online shopping with the haptic interaction features mentioned above. The result revealed users’ significant preference towards haptic-based product browsing over that provided by traditional online shopping regarding weight, geometry information and overall browsing satisfaction.
SCE YouTube link:
Mining review data to recommend doctors
Author: Huang Lingyun (Computer Science)
Supervisor: Associate Professor Cong Gao
There are a lot of online reviews on the medical treatments provided by the gynecologists. It will be too laborious for the human beings to process all the reviews at one go. Hence, this final year project is developed to analyze the comments in the web forums and find out the patients’ impressions on their gynecologists. Various measures have been explored to write an entity recognition algorithm in identifying the gynecologists being mentioned in the reviews. Data mining techniques have also been employed to dig out the essential information. Lastly, a web application is developed to let its users to know the relative cost of treatments and the experiences of the treatments with each gynecologist. Statistics on the Top 10 gynecologists in terms of medical experience and cost of treatment are also rendered on the web page.
SCE YouTube link:
Map-based Smart Property Search on Mobile Devices
Author: Tan Jia Hao (Computer Science)
Supervisor: Associate Professor Sourav Saha Bhowmick
With the growth of housing market and smart device users in Singapore, the property industry was slow to innovate as demand to search for properties efficiently on-the-go increases. This project aims to explore the inefficiencies introduced by property industry and develop a solution to better help consumers in making informed decision.
Existing property searching platforms limits consumer to traditional search methods such as search by location and search by property attributes. Property information are also inefficiently presented in a cluttered manner along with limited amenities details.
For this project, we have developed a mobile application that leverages on the availability of open source datasets on local infrastructures, as various map-based methodologies and APIs are studied and applied to build a smart map-based property search system.
The objective of the map-based search system is to analyse and link underlying interconnections among the data to provide beneficial insights to consumers that helps tackle their customised search requirements. The following are the proposed and implemented features of the search system:
1. Cluster-based Map View: Clustering of map markers
2. Smart Listing View: Informative listing of nearby amenities
3. Placed-based Search: Search for properties near a specific location
4. Theme-based Search: Search for properties near multiple types of amenities
5. School-based Search: Search for properties near a specific school
6. Fastest-route Search: Recommend properties based on shortest path to multiple destinations
This project has currently qualified for the semifinal of ACM CIKM AnalytiCup 2017, which is an open competition held in conjunction with ACM CIKM involving exciting data challenges aimed at the members of the industry and academia interested in information and knowledge management.
SCE YouTube link:
Author: Gupta Aakash (Computer Science)
Supervisor: Assistant Professor Zheng Jie
Scalable processing of large, heterogeneous, and possible incomplete and/or conflicting data, makes the analysis of haplotype data a challenging task. Moreover, near completion of the genome sequences and the re-focus on research analysis, makes the issue of effective genomic sequence display essential: it becomes cumbersome and difficult to understand to have billions of genomic DNA letters displayed on the screen as plain text! Thus, it is of paramount importance to be able to collect and digest the large amount of data about biological systems that is accumulating in the literature. Visualizing the data has successfully aided in gaining better understanding of the data. Moreover, researchers wish to view all facets of the genotype and haplotype data, including the spatial distribution of the loci along a chromosome, the different frequencies of haplotypes in different subgroups, and possibly also the correlation of occurring haplotypes. This emphasizes a need for a dynamic visualization which can address such complex and huge data sets on many different levels. As a solution, Singapore Immunology Network (SigN) aims to provide a customizable and highly user-interactive display of requested portion of genomes. Apart from kick starting the project, SIgN aims to release the project in the public domain to enable collaborators from all over the world to contribute to and expand the project.
As the foundational stone, three kinds of plots have been made to analyse genomic sequences in a better manner – Manhattan Plot, Genes Plot, and the Leaf Nodes Plot.
We believe with the help of collaborators all over the world, this visualization will aid the biologists in analysing DNA sequences and hence forming of hypothesis
SCE YouTube link:
Back to listing