In this paper, we report our ongoing research on a grammar formalism for retrieving information on the Semantic Web. We view the Semantic Web as a collection of databases and use a two-level grammar by which one can specify contextsensitive constraints that need to be satisfied. To retrieve information, a user can simply specify keywords and our system can show a result that is a string derived using the two-level grammar.
In this paper, we propose an approach to write
contextsensitive constraints using a grammar in order to retrieve
information on the Semantic Web. To this end, we use a
two-level grammar that is a 6-tuple (M, V, T, RM , RV , S),
where M is a finite set of metanotions, V is a finite set
of syntactic variables such that M ∩ V = ∅, T is a finite
subset of V +, RM is a finite set of metarules X → Y , where
X ∈ M , Y ∈ (M ∪ V )∗ or Y is a regular expression, and
for all W ∈ M , (M, V, RM , W ) is a collection of
contextfree grammar and regular expression rules, RV is a finite
set of hyperrules of the form H0 → H1, H2, · · · , Hm, where
m ≥ 1 and H0 ∈ (M ∪ V )+, Hi ∈ (M ∪ V )∗ for i ≥ 1, Hi
is a hypernotion, and S is a string of positive length over
M ∪ V , respectively [
The motivation of using a two-level grammar is that it allows to specify context-sensitive information in an intuitive way. Our approach allows a user to express facts that contain certain parameters which reflect structures of databases that constitute the Semantic Web. In other words, the formalism allows users to specify certain facts together with placeholders that can be instantiated using the data stored in databases.
The structure of this paper is as follows. Section 2 describes related research works. Section 3 explains the idea behind our approach using illustartive examples. Section 4 describes the structure of the system that we implemented. Section 5 concludes the paper and discusses research directions.
There are two research areas that are related to our
research. One is research about two-level grammar and the
other is retrieving information on the Semantic Web. Two
level grammar was introduced to define the syntax of
ALGOL 68 by van Wijngaarden [
In this section, we show how we can describe context-sensitive information using a two-level grammar. A string that can be derived using the grammar corresponds to certain information that can result from combining data that exist in the databases. There are two types of examples. The first example shows the case where we use one database that contains some number of tables. The second example shows how we can use two databases. Given this database, we can write metarules and hyperrules so that a string whose structure looks ’( ), ( ) travels ( ) with ( ) package in ( ) agency.’ can be derived as follows, where ( ) is a placeholder that can be instantiated with the data stored in the database tables.
I : : DATE, U ID U : : U ID U NAME t r a v e l s S ID S : : S ID S NAME with P ID P : : P ID P NAME package i n A ID A : : A ID A NAME agency .
where U ID i s U ID : t r u e where S ID i s S ID : t r u e where P ID i s P ID : t r u e where A ID i s A ID : t r u e Assuming that the tables contain data shown in table 1 to table 5, a possible derivation looks as follows.
ST ART ⇒ I U S P A ⇒ DAT E, U ID U S P A ⇒ 2012 − 7 − 2, 1001 U S P A ⇒ 2012 − 7 − 2, 1001 U ID U NAME travels S ID S P A
In order to show how the same approach can be used with multiple databases, we added a university database that consists of two tables. In addition, we modified the travel database. Customers in the travel database are students in the university database. Buy table contains the information about package purchasing for each customer and user name of travel database corresponds to student name of university database (figure 3). U NAME o f TRAVEL.NAME and U NAME o f USER : t r u e (3) The following hyperrules verify whether data is connected correctly in Univ DB.
D ID o f UNI .NAME i s same D ID o f DEPART i s same , D ID o f UNI .NAME and D ID o f DEPART : t r u e (4) The following hyperrules verify whether data is connected correctly in Travel DB.
P ID o f TRAVEL.NAME i s same P ID o f PACKAGE, P ID o f BUY i s same P ID o f PACKAGE, P ID o f TRAVEL.NAME, P ID o f PACKAGE and P ID o f BUY : t r u e (5) The following hyperrule verifies whether purchasing data is connected correctly in Travel DB. where U ID i s U ID : t r u e
Our system (TLG sysetm) has been implemented using Java
and HSQLDB system [
D_id 100 102 In this paper, we report our ongoing research on how a twolevel grammar can be used to retrieve information on the Semantic Web. We view the Semantic Web as a collection of databases and constraints existing in the colecction can be specified using hyperrules of the formalism that we proposed.
The motivation of the current research is that once we have
a declarative description about the Semantic Web using a
formal grammar, it becomes possible to process the Semantic
Web. In other words, a part of the Semantic Web can be fed
into a computer program as an input and the program can
parse the input and perform some task. This is in line with
the goal of utilizing the Semantic Web as a representation
medium for computations [
Currently, we are implementing a system that parses expressions defined using a two-level grammar and shows derivation results. We are also working on ways by which the information on the Semantic Web can be exploited using a