RESEARCH STATEMENT: Kuan-ming Lin, Feb. 5, 2006.
  
Since I took the Database course taught by Prof. Yang, I have been
attracted by the excitement of data mining. Broadly speaking, I am interested in
finding useful patterns in large stores of data that can be handled only by
computers. In earlier years, such large size of data was collected by 
business or organizations and kept in proprietary data warehouses.
Therefore, many data mining techniques were specialized to tackle
individual cases, and their generalizability has not been deeply studied. 
Today, as the Internet has become the largest information storage worldwide,
it is possible for the public to perform data analysis without access to
proprietary databases, so it is high time I participated in data
mining research. In particular, I am curious in mining two categories of
public-accessible electronic data, namely biological databases and Web documents.

In recent years, a number of repositories of biological data has been
published on the Internet, freely and publicly available to the global community.
Notably, the emerging of high throughput methods contributes to plentiful
public databases measuring different aspects of biological processes. The
data collectively are believed to provide more information than what they 
were explained in individual studies. For example, we could assign function
to unknown genes by looking at their expression correlations across multiple
microarrays. As a result, how to extract complementary information across
heterogeneous datasets is an appealing data mining problem for discovering
unknown biological relations. Since there is usually no underlying theoretical
model available for biological datasets, this type of data mining is very 
challenging and worth studying.

As the rapid growing of textual information on the Web, text mining on Web
documents is receiving much attention. That is, retrieving interesting
information from Web documents could have a high commercial potential
value. In academic research, Web text mining is a unique challenge over
traditional information retrieval in that Web documents are by nature
highly unstructured but interconnected through hyperlinks. Therefore, 
designing new techniques for mining Web document could lead to great
applications, and I will be happy to be involved in related research.

Throughout the last year, I started to develop skills in my interested data
mining subfields: My courses and initiation project for my second year 
focus on bioinformatics and data analysis, and I gained experiences on 
Web text mining through my research internship in Taiwan. In this year, I
hope to continue my research in data mining, and find a related research 
group in Duke who can support me throughout my PhD study.