Page 35 - profile2014.indd
P. 35
sensing systems: one is a mission-critical sensor network, called supporting communication and computation infrastructures for
YushanNet, for hiker tracking, search, and rescue in Yushan National gathering, caching, fusing, and distributing ubiquitous and het-
Park, and the other is a participatory sensing system, called TPE- erogeneous real-time streams of sensor data and information to
CMS, for comfort measurement of public transportation systems in response centers and individual responders and volunteers during
the Taipei metropolitan area. disasters; exploitation of complementary merits of different net-
work access technologies, approaches, and network types to make
Sheng-Wei Chen the physical connectivity as robust as possible during and after
disasters; and a combined named-data-networking and software-
Applications of human computation range from the exploitation defined-networking framework for enhancing disaster resilience of
of unsolicited user contributions, such as using tags to aid under- communication infrastructures.
standing of the visual content of yet-unseen images, to utilizing
crowdsourcing platforms and marketplaces (e.g., Amazon’s Me- Jan-Ming Ho
chanical Turk) to micro-outsource tasks (such as semantic video an-
notation) to a large population of workers. Further, crowdsourcing We are also interested in studying network computation problems
offers a time- and resource-efficient method for collecting large that arise when providing large-scale risk management services.
inputs for system design or evaluation purposes. We are applying Despite the long history of the development of economic and fi-
crowdsourcing to optimize computer systems more rapidly, and to nancial theories and practices in financial risk management, the
address human factors more effectively. In the past few years, we “worldwide credit crisis” in 2008 demonstrated the vulnerability of
have performed extensive studies on the performance of GWAP the current financial industry to risks. Several examples show that
(Games with A Purpose) systems and designed a human computa- even the three major rating agencies were unable to efficiently re-
tion game in order to efficiently collect diverse user annotations. port major default events. In Enron’s bankruptcy case in 2001, its
In addition, we have proposed a cheat-proof framework that can bonds maintained “investment grade” ratings until five days before
be used to effectively assess the quality of experience provided by the company declared bankruptcy. In Lehman Brother’s case in
multimedia content. We have found that crowdsourcing is indeed a 2008, they still received “investment grade” ratings on the morning
powerful strategy that can draw collective intelligence for AI-hard they declared bankruptcy. The rating companies claim that their
problems. In the future, we will continue to study how to effectively rating reports provide a long-term perspective rather than pro-
use crowdsourcing to overcome challenges in a variety of areas. viding an up-to-minute assessment. In rating credit of a company,
there are hundreds of firm-specific and macroeconomic variables.
Jane W. S. Liu There is no doubt that assessing credit risk in real-time is indeed a
task with high computational complexity. Nevertheless, it is an im-
A disaster management information system (DMIS) facilitates the portant foundation for maintaining stability of the financial market.
access, use, and presentation of data and information by applica- Complementary to the research into economic and financial theo-
tion systems and services that support decisions and operations ries and practices in risk management, we aim to develop comput-
during all phases of disaster management. State-of-the-art DMIS ing technologies for provision of large-scale, real-time financial risk
have several common limitations: they cannot make effective use management services, including (1) real-time rating of company
of all available information sources during emergencies; they do credit; (2) real-time rating of personal credits; and (3) pricing of fi-
not exploit synergistic information from networks of things and nancial products.
crowds of people; and they are not sufficiently agile in response
to changes in the disaster situation. We are collaborating with re-
searchers in the Institute of Earth Sciences and with engineering
faculty members from several leading universities in Taiwan and
the USA to develop an open framework for building a DMIS that
is free of these limitations. Our work is now supported by an Aca-
demia Sinica, Sustainability Science Research project called Ope-
nISDM (Open Information Systems for Disaster Management). The
current projects being undertaken by members of our laboratory
include the development of smart cyber-physical devices and ap-
plications as elements of a disaster-prepared smart environment;
algorithms and tools to support the collection and fusion of hu-
man sensor data contributed by crowds of people together with
surveillance data from in situ physical sensors; methods and tools
35