=Paper=
{{Paper
|id=Vol-1870/paper-05
|storemode=property
|title=Demonstration of Using a Domain-Specific Visual Modeler for Building Semantic Queries
|pdfUrl=https://ceur-ws.org/Vol-1870/paper-05.pdf
|volume=Vol-1870
|authors=Gábor Simon,Dániel Palatinszky,Gergely Mezei
|dblpUrl=https://dblp.org/rec/conf/esws/SimonPM17
}}
==Demonstration of Using a Domain-Specific Visual Modeler for Building Semantic Queries==
https://ceur-ws.org/Vol-1870/paper-05.pdf
Demonstration of Using a Domain-Specific
Visual Modeler for Building Semantic Queries
Gábor Simon, Dániel Palatinszky and Gergely Mezei
Department of Automation and Applied Informatics, Budapest University of
Technology and Economics
simon.gabor@aut.bme.hu, pd1326@hszk.bme.hu, gergely.mezei@aut.bme.hu
Abstract. In the age of Big Data, data mining, exploring and extracting
meaningful information from massive datasets became a natural need.
Currently this is a natural habitat for data scientists who have the
knowledge about standards, tools and query languages. While more and
more datasets were made available to the public, the exploration and
extraction tools for non-technical users remained limited. Usually only
simple keyword search or browsing through pre-determined navigation
paths are offered to this audience. The aim of our research is to create
an intuitive, user friendly visual query editing solution that is usable to
create queries for semantic web databases. We designed a domain-specific
language and built a fully customized, web-based editing workbench upon
it. The workbench achieves high usability and comprehensibility.
Keywords: semantic dataset, visual query, Wikidata
1 Introduction
It can be a challenging task to get the simplest piece of information from a
massive dataset without the support of a visual interface even for a user with
far-reaching domain knowledge. Data stores can be addressed in some kind of
specialized query language, but these languages require a not necessarily technical
user to master a highly technical language and translate his natural language
query to something that seems to be a cryptic and unnatural script for him. To
solve this issue, various kinds of higher level interfaces were developed as surveyed
recently in [7] in order to close the gap between the mindset of the user and the
comprehension of the data store [4]. All these interfaces are balanced between
expressiveness, flexibility and usability. Most of the highly usable tools restrict
user queries to match some of the predetermined query patterns [6], while the
other side of the spectrum is represented by highly technical editors visualizing
concepts and verbs directly from the underlying textual query language [1]. Our
research aims to raise the level of customization by offering greater flexibility in
expressing the queries, but at the same time keeping the visual language simple
enough to be understood by average, non-technical people.
Visual domain-specific modeling environments provide solution in a similar,
but more generic problem space. These environments offer a highly customizable
2 Gábor Simon, Dániel Palatinszky and Gergely Mezei
and flexible interface for arbitrary textual or visual domain-specific languages
(DSLs). The core of our solution is a metamodel-based, general purpose visual
domain-specific modeling system, the Visual Modeling and Transformation Sys-
tem (VMTS [2]). We have created a custom domain-specific language capable of
describing queries for the semantic data and customized the web-based editing
interface to meet the challenges in this domain. In this paper, we introduce our
editing environment referred to as SemEx (Semantic Extensions for VMTS) and
show its mechanisms by using an example.
2 The SemEx Visual Query Environment
The first step of our work was to create a language describing the queries. Since
our solution is based on a metamodeling system, the language is defined by a
metamodel. Our metamodel was created on the SPARQL query language [5] in
mind, however it does not capture the full expressive power of the SPARQL 1.1
grammar, as we focused our efforts to support the queries that can be expressed
using basic graph patterns (BGP).
In the SemEx metamodel the outermost concept is Query, which represents a
semantic query. A Query consists of Query Elements. There are several types
of Query Elements: Subject, Statement, Property and Object. Subject is a known
or unknown entity the users have and/or need information about. Statements
can be added to a Subject to encode the known or missing pieces of information.
A single Statement represents an atomic piece of information about a Subject,
and it contains a single Property and a single Object. The Statement st added
to Subject s with Property p and Object o encodes the knowledge that s has a
property p with the value of o. Moreover, we also distinguish Primitives that are
specialized Objects holding a literal value.
Query Elements can be either defined or undefined. Defined elements hold
a reference to a particular entity in the dataset or in case of Primitives, they
have a literal value. In contrast, undefined elements specify an alias that can be
referenced in the query results.
SemEx provides a web based interface to build a query in the form of a graph.
Nodes have customized visualization encapsulating a Subject and its Statements,
while an edge between two nodes encode the knowledge that both the nodes
denote the same entity or value.
To define a Query Element (e.g. Subject or Property), users are aided with a
simple filter interface that helps them to choose a concrete entity. The filter is
implemented as an auto-complete search box. In case of Primitives (literal values)
input fields are used. For undefined query elements, users can type in an alias. It
is possible to express equality between two elements by either using the same
alias, or by connecting the elements graphically. An edge between two subnodes
encode the knowledge that both the subnodes denote the same entity or value.
Once the visual query is built by the user, SemEx can generate a SPARQL
query from the model graph. The generated query can be inspected in a syntax
highlighted panel that also supports manual editing of the final query. Finally,
SemEx Visual Query Environment 3
the query can be executed from the environments on the specified SPARQL
endpoint and the results are visualized in a tabular form.
We tested our tool with one of the major publicly accessible semantic dataset
Wikidata [8] through utilizing its public SPARQL query service [3]. To test our
approach, we have built queries with complexity equivalent to the ones presented
in the user study of [6].
As a case study we present the visual model constructed to answer a question
with maximum complexity that users encountered through the user study. From
this type of question results a query with long (at least 3) linear chain of connected
elements and contains at least two undefined elements of the same type. Paper [6]
refers to this expressiveness category as "Long with branching and type III cycle
(T6)". We constructed the following category T6 task: “Find the names of all
the people that directed a science fiction movie and had a role in a film that has
an actor with a given name "George" who was born in the European continent
and plays in a rock band.” The resulting visual model is depicted in Figure 1. For
more models of the same case study, please refer to the SemEx page on [2].
Through the demo session we are going to build similar queries interactively,
driven by the input from the audience.
Fig. 1. Visual model of a T6-type sample query
3 Conclusions
Massive semantic datasets like Wikidata provide simple yet efficient tools for
everyone to browse and edit semantic data. However,even for the simplest queries,
4 Gábor Simon, Dániel Palatinszky and Gergely Mezei
users have to learn a textual query language. In this paper, we introduced
a new approach to create and execute visual queries on semantic datasets.
We recognized that crafting a query visually is essentially equivalent to the
process of editing a special visual model. As a result we were able to build our
tool on top of an existing visual modeling framework, VMTS. We mapped the
structure of SPARQL queries to a metamodel and thus created a domain-specific
language for this domain. Using the metamodel, VMTS provided an initial
editing environment, which we customized in order to improve usability. The
result, SemEx, offers a clean, accessible and uncluttered web interface to formulate
information retrieval tasks with ease. Moreover, harnessing the capabilities of
the underlying VMTS infrastructure, users can collaborate real-time through the
query formulation process, making SemEx a social data mining solution for people
without programming skills. Currently, the metamodel supports only queries
that can be translated to basic RDF graph patterns. We are working on adding
support for more complex query concepts, like group graph patterns or refined
relations (e.g. negation) between entities and values. As we built our approach
upon a visual modeling system, increasing the expressiveness is a natural process:
we enrich the metamodel and add more customization to the user interface. As a
result, we can offer a more capable tool with the continuous support of the base
services provided by the underlying modeling infrastructure.
Acknowledgments This work is connected to the scientific program of the
"Development of quality-oriented and harmonized R+D+I strategy and functional
model at BME" project. This project is supported by the New Széchényi Plan
(Project ID: TÁMOP-4.2.1/B-09/1/KMR-2010-0002).
References
1. OpenLink iSPARQL, https://www.openlinksw.com/isparql
2. VMTS website, http://vmts.aut.bme.hu
3. Erxleben, F., Günther, M., Krötzsch, M., Mendez, J., Vrandečić, D.: Introducing
Wikidata to the linked data web. In: Proceedings of the 13th International Semantic
Web Conference (ISWC’14). LNCS, vol. 8796, pp. 50–65. Springer (2014)
4. Freitas, A., Curry, E., Oliveira, J.G., O’Riain, S.: Querying heterogeneous datasets on
the linked data web: challenges, approaches, and trends. IEEE Internet Computing
16(1), 24–33 (2012)
5. Harris, S., Seaborne, A., Prud’hommeaux, E.: Sparql 1.1 query language. W3C
recommendation 21(10) (2013)
6. Soylu, A., Giese, M., Jimenez-Ruiz, E., Vega-Gorgojo, G., Horrocks, I.: Experiencing
OptiqueVQS: A multi-paradigm and ontology-based visual query system for end
users. Univers. Access Inf. Soc. 15(1), 129–152 (2016)
7. Vega-Gorgojo, G., Slaughter, L., Giese, M., Heggestøyl, S., Soylu, A., Waaler, A.:
Visual query interfaces for semantic datasets: An evaluation study. Web Semantics:
Science, Services and Agents on the World Wide Web 39, 81–96 (2016)
8. Vrandečić, D., Krötzsch, M.: Wikidata: A free collaborative knowledgebase. Commun.
ACM 57, 78–85 (2014)