=Paper=
{{Paper
|id=Vol-295/paper-17
|storemode=property
|title=SMART: A Web-Based, Ontology-Driven, Semantic Web Query Answering Application
|pdfUrl=https://ceur-ws.org/Vol-295/paper17.pdf
|volume=Vol-295
|authors=Alexander De Leon Battista,Natalia Villanueva-Rosales,Myroslav Palenychka,and Michel Dumontier
|dblpUrl=https://dblp.org/rec/conf/semweb/BattistaVPD07
}}
==SMART: A Web-Based, Ontology-Driven, Semantic Web Query Answering Application==
https://ceur-ws.org/Vol-295/paper17.pdf
SMART: A Web-Based, Ontology-Driven, Semantic
Web Query Answering Application
Alexander De Leon Battista1, Natalia Villanueva-Rosales1, Myroslav Palenychka1,
Michel Dumontier1,2
1
School of Computer Science, 2Department of Biology,
Carleton University, 1125 Colonel By Drive, Ottawa, Ontario, K1S5B6 Canada
{adlbatti,nvillanu,mpalenyc}@scs.carleton,ca, michel_dumontier@carleton.ca
Abstract. SMART (Semantic web information Management with automated
Reasoning Tool) is an open-source project, which aims to provide intuitive
tools for life scientists for represent, integrate, manage and query heterogeneous
and distributed biological knowledge. SMART was designed with
interoperability and extensibility in mind and uses AJAX, SVG and JSF
technologies, RDF, OWL, SPARQL semantic web languages, triple stores (i.e.
Jena) and DL reasoners (i.e. Pellet) for the automated reasoning. Features
include semantic query composition and validation using DL reasoners, a
graphical representation of the query, a mapping of DL queries to SPARQL,
and the retrieval of pre-computed inferences from an RDF triple store. With a
use case scenario, we illustrate how a biological scientist can intuitively query
the yeast knowledge base and navigate the results. Continued development of
this web-based resource for the biological semantic web will enable new
information retrieval opportunities for the life sciences.
Keywords: Semantic Web, Query Answering, Graphical User Interface.
1 Introduction
The primary goal of the Semantic Web is to add semantics to the current Web in such
a way that machines can reason about the content [1]. Research activities in the
Dumontier Lab are focused on using semantic web technologies to enable
biochemical knowledge discovery [2, 3]. In combination with complimentary efforts,
we expect that more powerful knowledge discovery tools will facilitate data
integration in bioinformatics and create powerful new data mining opportunities over
heterogeneous biochemical knowledge. However, friendly interfaces to query over the
semantic web remain a critical limitation in their use and adoption.
The open-source Semantic web information Management with automated
Reasoning Tool (SMART) project aims to provide intuitive interfaces for the
representation, integration, management and querying of knowledge using the
semantic web. In this paper, we describe an initial web-based application that adopts
an ontology centric model to perform semantic query answering over heterogeneous
yeast biological knowledge. Ontologies in this application are described using OWL
DL, a sublanguage of the Web Ontology Language (OWL) [4]. Features of the
application include semantic query composition and validation using DL reasoners, a
graphical representation of the query, a mapping of DL queries to SPARQL 1 and the
retrieval of pre-computed inferences from an RDF triple store.
2 System Architecture
The system architecture shown in Fig. 1 is composed of several components including
an application mediator, a web-based user interface, an ontology repository, an
ontology index, a hybrid reasoner which optimizes the reasoning process by
converting a DL expression into SPARQL, and a SPARQL query engine for querying
a database/triple store. SMART was implemented using the Java programming
language and interoperates with other open source and Semantic Web technologies,
including the WonderWeb OWL API 2 , Pellet 3 , Jena 4 , and Protégé 5 .
Fig. 1. SMART Architecture
SMART Mediator: The SMART Mediator is the core component of the system
which enables decoupled interactions among components, controls the behavior of the
system, and manages ontologies used by the application. The mediator loads URI-
identified ontologies via the ontology repository. The retrieved ontology becomes the
active ontology upon which reasoning and query answering may occur. The mediator
will also resolve and load ontologies imported by the active ontology. The mediator
enables communication between components and this supports sophisticated multi-
component behavior such as the use of a reasoner or ontology index by the user
1 http://www.w3.org/TR/rdf-sparql-query/
2 http://wonderweb.man.ac.uk/owl/
3 http://pellet.owldl.com/
4 http://jena.sourceforge.net/
5 http://protege.stanford.edu/
interface. The manipulation of OWL documents and data structures by multiple
SMART components is facilitated by the OWL API.
Ontology Repository: The ontology repository supports storage and retrieval of URI-
identified ontologies. The current implementation will store new or modified
ontologies into the repository. Ontologies can be stored and retrieved in/from the
repository by using its URI or the combination of the URI and a version number, thus
it provides a simple mechanism for versioning modified ontologies. This
implementation uses a file based system to store ontologies, but other
implementations could use more sophisticated storage solutions such as relational
databases or version control systems.
Ontology Indexer: The ontology indexer enables fast and easy access to entities
defined in the active ontology. Indexed entities include class, property, individual
names and their types. Thus, this component supports partial entity name matches,
and retrieval of specific types of entities. The current implementation leverages the
speed and power of the Lucene 6 Text Indexer, therefore, it eliminates the need for in-
memory data structures for caching entity metadata.
Hybrid Reasoner: Reasoning about large ontologies can be expensive both in terms
of time and space. For this reason, we have implemented a “cached-reasoning”
service that retrieves pre-computed inferences about an OWL-DL ontology.
Ontologies loaded into the system are sent to a DL reasoner offline and the inferences
retrieved are explicitly stored in the triple store. The Hybrid Reasoner component uses
the DL2SPARQL component to convert an OWL Class expression to a SPARQL
query which is sent to a SPARQL query engine via the SMART mediator. The hybrid
reasoner currently answers queries that involve subsumption, realization and role
restrictions. Queries composed of negation or cardinality restrictions currently require
a DL reasoner, or a knowledge base that integrates a DL reasoner. Nevertheless, this
component optimizes the querying of large OWL ontologies that could take a long
time to reason about or would require more memory than is currently available.
SPARQL Query Engine: The SPARQL Query Engine provides the ability to query a
triple store repository using SPARQL. The interface provides an API for components
that require SPARQL query services, such as sending a SPARQL query and obtaing a
result set. Implementations for Jena, AllegroGraph and Sesame serve as wrappers for
the vendor-specific APIs, hiding them from the components that use the services.
Graphical User Interface: SMART’s web-based, graphical user interface supports
semantic query construction and validation, result retrieval and presentation. The
graphical interface is built primarily on the combination of two important
technologies: Java Server Faces (JSF) 7 and Asynchronous Javascript and XML
(AJAX) 8 . This combination supports the development of reusable event-driven UI
6 http://lucene.apache.org/
7 http://java.sun.com/javaee/javaserverfaces/
8 http://ajax.asp.net/
components connected to a server application that eliminates the need to reload pages.
Such behavior was desirable as certain components were expected to be re-used by
different parts of the user interface, and in which the ability to react to user content in
real time was also deemed essential. The user interface has been tested and operates
optimally with Mozilla FireFox 1.5.
3 Semantic Query Answering
We believe that semantic query answering occurs when ontologically-defined entities
and roles are used to answer conjunctive queries. In this way, ontologies define the
lexical granularity from which queries may be formulated and offer the necessary
semantics to evaluate whether an answer to a query may possibly exist.
A query language will determine the possible questions that a user can ask to the
system. For example, in the relational database world, SQL is the standard language
used for creating queries. Several rich query languages exist for querying RDF/OWL
data. These include RQL [5, 6], nRQL [7]. While powerful queries may be formulated
using these languages, they require training and thus pose a challenge for non-experts
users. To facilitate semantic query formulation, we utilize the Manchester OWL
Syntax [8] to define DL queries (i.e. queries defined by a class description). An
important feature of this language, which makes it more readable and easier to
understand than traditional descriptive logics syntax, is that it uses an infix notation
rather than a prefix notation for keywords. E.g. the traditional DL expression: Father
! ! isMarriedTo.Architect would be written as “Father that isMarriedTo some
Architect” in this syntax. While room for improvement still exists, constructing class
expressions using this syntax is substantially easier to formulate by non-DL experts,
and reduces the technical barrier of learning another query language.
In SMART, DL queries are created in a text box using the web-based user
interface. Real-time feedback allows users to construct syntactic, semantic and logical
valid queries. This feedback is given as follows: First, a list of entities contained in
the knowledge base (class, property, and individual names) are suggested based on a
partial word matches as they are entered by the user. The user may continue typing
the word or select one of the suggested words. However, if the user types a word that
is not found in the system, an error appear as the query is not valid. Moreover, the
correct type of entity is suggested for the phrase under construction thereby
encouraging users to build grammatically valid queries. Second, a valid DL query is
verified by a DL reasoner on the server side, and a message appears below the query
expression which notifies, as the user enters the query, if his/her class expression is
valid or not. This allows early recognition of semantic errors, and enforces logically
valid expressions. Finally, a graphical representation of the query is displayed to aid
in the interpretation of complex or nested expressions. These dynamic features were
made possible by leveraging the Manchester OWL Syntax parser from Protégé 4
(alpha) that convert the DL query into an OWL API object that can be exchanged
with a DL reasoner.
Use Case Scenario: Querying Yeast Biological Knowledge
This section aims to illustrate how SMART may be used for semantic query
answering over yeast biological knowledge. We loaded the yOWL ontology and
3,423 instances populated from the Saccharomyces Genome Database (SGD). For
more information about this ontology, refer to [2]. The yOWL ontology currently
contains 244 classes, from which 38 are defined, 57 object properties and 22
annotation properties 9 .
In this scenario, a scientist wants to know which yeast genes/proteins that exhibit
transferase activity (aka transferases) are part of a molecular complex and have been
associated with a nonviable (or fatal) outcome when removed.
Our scientist will first access SMART online at http://smart.dumontierlab.com and
begin composing her query in the query text box. She first tries “transferase”, but
SMART offers no matching terms in the suggestion box. She decides to make her
query more explicit by defining exactly what she means. She starts by typing “Gene”,
SMART suggests this word among others based on partial word matching to
knowledge base entities. She selects “Gene” by hitting the tab button, presses the
space bar, and begins typing a character. A message indicates that the expression is
not valid and when she presses the “why?” link, a window appears to explain that
SMART expects a boolean class constructor (and, or, not, that) to formulate a
conjunctive query. She quickly types the “and” operator.
Fig. 2. SMART web-based interface for query composition with suggestions.
Next, she would like to look for genes that have the transferase function as defined
by the Gene Ontology (GO) 10 . Fig. 2 illustrates the property suggestions starting with
the letters “has”. She selects “hasFunction” and is now prompted to select
existential/universal or cardinality restrictions (some, only, value, min, max, exactly)
to modify the range of the predicate. Knowing that genes/proteins often have multiple
functions, she wants to find “some” that have at least this function and defines the
property range as the GO term Transferase_activity. She formulates the following
syntactically, logically and semantically correct DL query to her question:
Gene and hasFunction some Transferase_activity and isPartOf some Complex and
isParticipantIn some (Experiment and hasOutcome some Nonviable)
9 http://ontology.dumontierlab.com/yOWL-1.0
10 http://www.geneontology.org/
A corresponding graphical representation of the text query can simultaneously be
seen in the Query Graph tab (Fig. 3). This graph query will allow our scientist to
visualize the query in a more intuitive way. Once the user is satisfied with the query,
she can then click the Answer button to obtain the results.
Fig. 3. SMART web-based interface screenshot showing the synchronization between
the text query and the query graph generated.
This query will now be answered by the hybrid reasoner which maps the DL query
to the following SPARQL query to be answered, in this case, by Jena:
SELECT DISTINCT ?i
WHERE {{{
{?i _:var2.
_:var2
.
_:var2 _:var3.
_:var3
}}
{?i _:var0.
_:var0
.
?i _:var1.
_:var1
.
?i
}}}
The result is then displayed in the Results tab (Fig. 4), which supports HTML
navigation of RDF/OWL graphs by dynamically retrieving individuals (nodes) their
associated properties (edges) from the knowledge base. For instance, on selecting the
“BET4” protein, a new panel to the right of the one selected lists properties associated
to “BET4”. On selecting the “isParticipantIn” property, a new panel opens to the right
of the property panel. By selecting one of the results such as “experiment_30”, she
may further explore what is know about “experiment_30” or backtrack to explore
other individuals. Thus, the path across multiple related individuals can be kept while
providing the flexibility to investigate other individuals and their relations.
Fig. 4. SMART web-based interface screenshot showing the results navigation.
4 Discussion
SMART empowers users to create simple expressions that allow them to pose
sophisticated conjunctive queries over a knowledge base. Query answering operates
under the open world assumption and the lack of unique name assumption according
to the Semantic Web semantics. The results obtained from these queries may include
class membership inferences (using necessary and sufficient conditions to define
classes) that are not available from browsing the original SGD database. Moreover,
capturing user-generated definitions (i.e. transferase as defined in the use case) will
create opportunities for capturing ontological concepts with more powerful semantics
than with Web 2.0 tagging. SMART can handle any ontology in the OWL language
for semantic query answering. To overcome problems with on the fly reasoning about
a large knowledge base, we used pre-computed inferences using a DL reasoner. This
substantially speeds up query answering on typical queries involving class
subsumption and class membership over a large amount of distributed and
heterogeneous biological data. However, substantial challenges remain for truth
maintenance upon the addition of new knowledge.
A general design principle of the user interface, based on results obtained from a
recent survey carried out in our lab for Canadian scientists, was to investigate the use
of English-like phrases for query composition. Surveys on the ease of use and
functional aspects by biology and computer science students in the lab have yielded
positive feedback, although a more rigorous usability study has yet to be performed.
While the user-tested Manchester OWL syntax and the Protégé expression parser lend
a first step towards this directly, it is anticipated that other, more sophisticated
controlled natural languages such as Sydney OWL syntax [9] will play a major role as
an enabling technology. In this work, we also explored the use of a graphical
language (similar to GLOO [10]) for simplifying the interpretation of complex
(and/or) nested expressions. We expect that the addition of interactive elements to our
graphical graph query interface will enable an interesting and alternative mechanism
for query formulation that might appeal to a large segment of the public. Since a user
never has to read/write/understand XML/RDF/OWL, SPARQL, SQL or any other
query language created for machine consumption, we consider that further
developments in this area is essential for non-experts to tap into the potential of the
semantic web.
5 Conclusions and Future Work
In this paper, we described the modular architecture and basic functionality of
SMART, an easy to use web-based application for the semantic web. Initial
functionality includes semantic query answering and browsing the results over an
OWL knowledge base using (when possible) pre-computed inferences. Ongoing work
includes exploratory content browsing, interactive graphical query formulation,
knowledge composition, and truth maintenance. Efforts are also being directed to
support user-specified ontology imports, queued reasoning, and implementation of
DL extensions for SPARQL.
Acknowledgments. We would like to thank Kirill Levinsky, Faris Matari and Anne Taylor for
technical contributions and useful discussions. This work was made possible by Luc Lalande
and funding from Foundry Program for AL and Talent First for MP. Funding for NVR and
MD was with CONACYT scholarship #150581 and NSERC, respectively.
References
1. Berners-Lee, T., Hendler, J., Lassila, O.: The Semantic Web: A new form of Web content
that is meaningful to computers will unleash a revolution of new possibilities. (2001)
2. Villanueva-Rosales, N., Osbahr, K., Dumontier, M.: Towards a Semantic Knowledge Base
for Yeast Biologists. HCLS-DI (2007)
3. Villanueva-Rosales, N., Dumontier, M.: Describing chemical functional groups in OWL-DL
for the classification of chemical compounds. OWLED (2007)
4. W3C: OWL Web Ontology Language Guide. In: Smith, M.K., Welty, C., McGuinness, D.L.
(eds.) (2004)
5. Haarslev, V., Möller, R., Wessel, M.: Querying the Semantic Web with Racer + nRQL. KI-
04 Workshop on Applications on Description Logics, Ulm (2004)
6. Gregory, K., Sofia, A., Vassilis, C., Dimitris, P., Michel, S.: RQL: a declarative query
language for RDF. Proceedings of the 11th international conference on World Wide Web.
ACM Press, Honolulu, Hawaii, USA (2002)
7. Fikes, R., Hayes, P., Horrocks, I.: OWL-QL--a language for deductive query answering on
the Semantic Web. Web Semant. Vol. 2 (2004)
8. Horridge, M., Drummond, N., Goodwin, J., Rector, A., Stevens, R., Wang, H.H.: The
Manchester OWL Syntax. OWLED 2006, Athens, Georgia (2006)
9. Fuchs, N.E., Schwitter, R.: Web-Annotations for Humans and Machines. ESWC (2007)
10. Fadhil, A., Haarslev, V.: GLOO: A Graphical Query Language for OWL Ontologies.
OWLED 2006 (2006)