![](bg2.png)
revising the accumulated cases on a case-by-case
basis. Second, when we use cases, we have less
information loss than when we use rules or frames.
Cases can include context information that contains
the tacit knowledge within documents, and it will be
helpful to maximize the information users’
satisfaction.
In the customization phase, we generate new
information that satisfies the users’ requirements by
using case-based reasoning. The proposed case-
based reasoning uses the two-step similarity
calculation method. In step 1, the similarity between
the information that the users’ requires and the
features of the cases is calculated by using the
requirements-feature similarity method. In step 2,
the requirement-value similarity is used to calculate
the similarity between the details of values and the
users’ requirements.
The proposed case is structured by features and
values. A value is defined by sentences that include
features. Therefore, a feature can be matched into
several sentences as values. To find a similar case,
we filter the cases according to the similarity
between case and the information of the user’s
requirements. Ontology is needed to determine the
semantic relationship between the features or values.
The selected case through the two-step similarity
calculation can be adjusted according to the user’s
requirements, which is conducted by the Case
Customization Agent.
In chapter 2, after review related studies, in
chapter 3, the overall architecture of the proposed
system is introduced. Next, we discuss the case and
the sub-case generation, and we describe how to
retrieve similar cases with using the two-step
similarity calculation. In chapter 6, we introduce the
case customization agent. Finally, we discuss future
study and the conclusion of this study.
2 LITERATURE REVIEW
Information integration is the process of
semantically and syntactically combining the data
that comes from different resources with adhering to
a unified format (Bernstein A. Philip and Laura M.
Haas, 2008). To do this, it is important to classify
the information resources such as the Web
documents into structured or unstructured forms.
The former are needed for transforming the data,
which is obtained from databases with the
heterogeneous schemas, into homogeneous data with
using a semantically and syntactically. This can be
accomplished by the ETL technology or the
mediated schema (Halevy Y. Alon, et al., 2005;
Bernstein A. Philip and Laura M. Haas, 2008). The
latter is able to extract information from
unstructured document through keyword search,
summary data, and pre-defined information
structures (e.g., the customer’s name and address,
the product information and the brand name, and so
on). An annotation such as the Resource Description
Framework (RDF) can also be applied to the
information selection and integration processes. The
process of applying ontology to the semantic
integration of information is getting more popular
regardless of the structure of that information
(Alexiev Vladimir, et al., 2005).
However, It is too difficult to integrate information
by using the selected documents based on a keyword
search due to the increase of web documents from
around the world by geographic progression. Even
we use the RSS, we can still receive the selected
information; so, the information users have to have a
part in electing the appropriate information. In
addition, it is necessary to check and resolve
possible conflictions. Information users usually
spend a lot of time and effort to get the customized
information (Mani, I. and Bloedorn, E., 1998; Teufel,
S. and Moens, K., 1997.). Therefore, it is critical to
propose an adequate methodology of information
integration. However, information is basically
differentiated from physical products (Lee, Y., et al.,
2001; Karmarkar, U. S. and Apte, U. M., 2007).
Therefore, to integrate information, research is
needed to understand the contents in documents, to
classify those contents and to integrate the contents.
To accomplish this, we suggest an Agent-based
Information Customization System with using the
CBR and Ontology.
3 OVERAL ARCHITECTURE OF
AGENT-BASED INFORMATION
CUSTOMIZATION SYSTEM
In an Agent-based Information Customization, Case
Generation Agent is for transforming the collected
information into cases, and Case Customization
Agent is for customizing the information depending
on the information of the user’s requirements. Their
overall architecture is shown in Figure 1.
ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence
286