Database Reference
In-Depth Information
Figure 12. Basic database searching process
Note that all afore-mentioned approaches (BANKS, DBXplore and DISCOVER) are useful to users
who do not know SQL or are unfamiliar with the database schema. However, they present a semantic
challenge because the metadata of attributes and relations that are in an SQL statement are lacking in
keyword search. Furthermore, because a keyword may appear in any attributes and in any relations, the
result set may be large and include many answers users do not need.
2.4 Discussion
Database search processes basically involve two steps, namely query formulation and query evaluation
(Figure 12). In the query formulation step, the user formulates her/his information need (or
retrieval
goal
) in terms of an SQL query. The query evaluation step runs this query against the database and
returns data that match it exactly.
Thus, formulating a query that accurately captures the user's retrieval goal is crucial for obtaining
satisfactory results (i.e., results that are both useful and of manageable size for human analysis) from a
database. However, it is challenging to achieve for several reasons:
•
users may have
ill-defined
retrieval goals, i.e., they do not really know what might be useful for
them, merely an expectation that interesting data may be encountered if they use the database
system (e.g.,
'I can't say what I want, but I will recognize it when I see it'
) ;
•
even if users have
well-defined
retrieval goals, they may not know how to turn them into regular
SQL queries (e.g.,
'I know what I want, but I don't know how to get it'
). This may be either because
they are not familiar with the SQL language or with the database schema. It may also be due to an
expressiveness limit of SQL. In fact, an information need is a mental image of a user regarding
the information she/he wants to retrieve and it is difficult to capture it using an unnatural exact
language such as SQL.
Note that even if users have
well-defined
retrieval goals and know how to formulate them in terms of
SQL queries (e.g.,
'I know what I want and I know how to get it'
), they may not obtain what they need
(e.g.,
'I am not satisfied'
). This occurs if the database they wish to access contains no data that satisfy
their retrieval goals.
One can clearly notice that the
Many-Answers
and the
Empty-Answer
problems are immediate con-
sequences of the above problems. In fact, if the user has an
ill-defined
retrieval goal, her/his queries are
often very broad, resulting in too many answers; otherwise, they are very specific and often return no
answers.
