Using Concepts in Grid Data Organization - Advanced Research on Computer Education, Simulation and Modeling - page 450

4 Content Search in Grid
Processing a linked pattern searching in grid file includes structural search and con-
tent search. Most existing algorithms do not differentiate content search from struc-
tural search. They treat content nodes the same as element nodes during searching
processing with structure. Due to the high variety of contents, to mix content search
and structure search suffers from organization problem of contents and low perform-
ance. Another disadvantage is to find the actual values asked by a searching, they
have to rely on the original file. Therefore, we propose a novel algorithm Value Ex-
traction with Relational Table (VERT) to overcome these limitations. The main tech-
nique of VERT is introducing relational tables to store file contents instead of treating
them as nodes and labeling them. Tables in our algorithm are created based on con-
cept information of files. As more concepts captured, we can further optimize tables
and queries to significantly enhance effciency.
For example, consider the example grid tree with containment labels. Instead of
storing the label streams for each grid tag and value contents, we can store the value
contents together with the labels of their parent tags in relational tables. With these
relational tables, when a user issues a linked searching , the system can automatically
rewrite it to the searching, where the node price, the value node with value greater
than 15, and their PC relationship are replaced by a node called price'>15. Then, we
can execute SQL in table Rprice to get all labels of price elements with value greater
than 15 to form the label stream for price'>15; and perform structure based on label
streams of book, ISBN and price'>15. In this way, we save both the stream merge
cost of all content values greater than 15 and the structure between the merged label
streams for content values and price element.
Moreover, if we know that price is a property of book object class by exploiting the
mode information, we can directly put the value contents of price with labels of book
object class, instead of the labels of price element. In this way, when processing the
linked searching, we can also save the structure between book object and its price
property. Note that we can also store the labels of book objects with the contents of
other properties, such as title, author, etc., which are not shown due to limited space.
Finally, if we further know that ISBN, title, price, etc. are single valued properties
of book object class according to concepts captured by ORASS, we can premerge the
content values of these properties into a single relational table with the labels of book
objects.With the premerged table, to answer the linked searching, we can simply per-
form an efficient selection on the pre-merged table without time consuming structure.
Note that we should not merge multi-valued properties (e.g.author) into the table to
avoid duplicate information.
5 Keyword Search with Concepts in Grid
Keyword proximity search is a user friendly way to searching grid databases. Most
previous efforts in this area focus on keyword proximity search in grid based on either
a tree data model or a graph (or digraph) data model. Tree data models for grid are
generally simple and efficient. However, they do not capture connections such as ID
references in grid databases. In contrast, techniques based on a graph (or digraph)