Database Reference
In-Depth Information
Now suppose that the user poses her/his query to a real estate agent. The estate agent often provides
that user with better results than the ones obtained using traditional database systems, as she/he has a
large amount of knowledge in the field of house buying. Let us briefly discuss two important features
of the estate agent that contribute to her/his ability to fit the user's information requirements:
1. the first is her/his ability to organize, abstract, store and index her/his knowledge for future use. In
fact, besides organizing her/his knowledge into groups (Miller, 1962; Mandler, 1967), the estate
agent stores (in her/his memory) these groups as a knowledge representation and such groups often
become an automatic response of her/him. This interesting statement comes from the central tenet of
semantic network theory (Quillian, 1968; Collins & Quillian, 1969), which argues that information
is stored in human memory as a network of linked concepts (e.g., the concept 'house' is related by
the word 'is' to the concept 'home'). Moreover, psychological experiments (Miller, 1956; Simon,
1974; Halford, Baker, McCredden & Bain, 2005) show that humans can deal with a large amount
of information, exceeding their memory limitations, when such information are supplemented with
additional features such as a relationship to a larger group (or concept). Hence, cognitive theories
assume that humans arrange their knowledge in a hierarchical structure that describes groups at
varying levels of specificity. Furthermore, Ashcraft (Ashcraft, 1994) found that humans assign
meaningful words from natural language to groups and retrieve information by those words (i.e.,
group representatives) rather than blindly traversing all information;
2. the second feature is her/his ability to assist the user to refine and clarify her/his information need as
well as to make a decision. In fact, the estate agent establishes a dialog with the user during which
she/he asks pertinent questions. Then, for each user's response (i.e., a new information need), the
estate agent uses her/his knowledge to provide the user with concise and comprehensive information.
Such information is retrieved by matching user query words with group representatives (Ashcraft,
1994) stored in her/his memory.
From these cognitive and technical observations, we propose a simple but useful strategy, that emulates
the interaction a user might have with a real estate agent to some extent, to alleviate the two previously
mentioned problems (i.e.,
relevance
and
scalability
). It can be summarized as follows (see Figure 14):
Figure 14. Our proposal
