=Paper=
{{Paper
|id=Vol-2451/paper-32
|storemode=property
|title=Visual Query Environment over RDF Data
|pdfUrl=https://ceur-ws.org/Vol-2451/paper-32.pdf
|volume=Vol-2451
|authors=Kārlis Čerāns,Julija Ovcinnikova,Lelde Lace,Jūlija Hodakovska,Aiga Romane,Mikus Grasmanis,Elina Kalnina,Arturs Sprogis,Agris Sostaks
|dblpUrl=https://dblp.org/rec/conf/i-semantics/CeransOLHRGKSS19
}}
==Visual Query Environment over RDF Data==
https://ceur-ws.org/Vol-2451/paper-32.pdf
Visual Query Environment over RDF Data
Kārlis Čerāns1, Jūlija Ovčiņņikova, Lelde Lāce, Jūlija Hodakovska, Aiga Romāne,
Mikus Grasmanis, Elīna Kalniņa, Artūrs Sproģis, Agris Šostaks
Institute of Mathematics and Computer Science, University of Latvia
1
karlis.cerans@lumii.lv
Abstract. We demonstrate the interactive ViziQuer environment for composing
rich visual queries (including aggregation, nesting and data expressions) over
SPARQL endpoints. We describe the interactive means for building the visual
constructs of the query language and discuss the user study results on the query
environment usability. The ViziQuer tool environment is publicly available and
open source.
Keywords: Visual queries, query environment, ad-hoc queries, RDF, SPARQL
1 Introduction
Visual query composition paradigm (cf. [1], [2], [3], [4], [5]) along with facet-based
([6], [7]) and controlled natural language based ([8]) approaches offers a promising
avenue for involving end-users in query composition over SPARQL endpoints (cf. [1]).
The recent ViziQuer notation ([5], [9]) allows for visual presentation of rich instance
level and aggregate queries, involving data expressions, as well as query nesting, with
expressive power approaching that of the full SPARQL 1.1 [10]. The corresponding
query construction environment reported so far [11] has been rather basic with hierar-
chic property pane of the visual elements as the primary query building tool.
We shall report in this paper and present in the demonstration:
Interactive rich visual query building environment with class tree basis and
context-based query construction options;
Context-dependent code completion of attribute and condition expressions and
property paths;
Ready-to-use data schemas over a number of existing public Linked data end-
points (the query examples in this paper are over Scholarly data endpoint [12]).
The user studies identifying the strengths as well as remaining areas of necessary
attention within the visual query creation environment are discussed, as well.
There are interactive multi-modal visual query creation environments available in
Optique VQs [1] and LinDA [3] visual query systems. The ViziQuer system considered
here is meant for much wider visual query range (including aggregation and subqueries,
as well as textual language for conditions and attribute expressions).
The open source ViziQuer query environment and its supporting resources are avail-
able from the home page http://viziquer.lumii.lv/, including the tool local set up options.
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
2 Notation Examples
The visual queries are formulated in the context of a data dictionary that is a part of the
data schema used by the ViziQuer tool (cf. Section 3), the data dictionary maps the
entity short names used in the visual query presentations to their full IRIs.
Each query is a connected graph with one main query node (orange round rectangle).
The query examples in Table 1, formulated over Scholarly Data schema [12], show a
class-attribute-link-condition option (1), queries with aggregation and grouping (2 and
3), queries with nested aggregation (4 and 5) and queries with advanced structure (6
and 7). The ViziQuer notation is explained in detail in [5] and [9] and the ViziQuer tool
home page. The new notations here, if compared to [9], involve explicit grouping (3),
instance IRI references (4 and 7) and distinct values concrete syntax (5).
Table 1. Visual Query examples
1. List all non-paper-related roles
during a conference starting in 2017,
together with the role holder names.
Each query node is a data instance
pattern, listing its class name and con-
ditions, and the selection items; the
links show the instance connections.
2. Find the number of conferences
and the number of conferences by the
year of their start date (two queries).
The grouping is automatic by all non-
aggregated selection fields.
3. For every keyword find its usage
count (sort descending) in papers
within each conference. The keyword
presence is mandatory, denoted by
{+}. The Conference instance is ex-
plicitly added to the grouping set.
4. List titles of all papers from the
iswc 2018 conference, together with
their author count (sort descending).
Nested query construct (dc:creator
link) is used for calculating author
count for each paper separately.
5. List titles of all papers together
with the papers’ first author name(s).
The data may contain duplicate rela-
tion information. So, the distinct per-
son name values are selected for each
paper, then concatenated to include
all name forms in the output.
6. For every conference find the num-
ber of persons that are either authors
of its proceeding papers or have some
role during the conference.
The visual links from the union node
[+] describe the data links from the
Conference instances.
7. Find the keywords used in proceed-
ing papers of both iswc and eswc con-
ferences in 2017, with respective us-
age counts, order descending by the
usage count in iswc 2017.
The unit node [] is a wrapper around
the two subquery results (there can be
combining, filtering, or even further
aggregation within wrapper nodes).
3 Query Environment
The query environment serves the creation of queries (such as the ones listed in Table
1) over a knowledge graph schema describing the structure of the data to be analyzed.
The query environment setup requires loading the data schema, listing the available
entity names and connections. The data schema can be extracted either from an OWL
ontology or directly from the SPARQL endpoint. There is a list of pre-built schemas,
as well as the reference to the schema extraction service available at the ViziQuer
Schema Store1. The Scholarly data schema used in the Section 2 examples has been
retrieved into the Schema Store from the Scholarly Data SPARQL endpoint2. There is
also an example project with the Scholarly data queries available in the Schema Store.
The reference to a SPARQL endpoint is to be specified within a project (in the pro-
ject parameter window), if the queries are to be directly executed over the endpoint.
The query creation process typically starts by double clicking a node in the class tree
(Fig. 1, left) thus adding a query node (orange round rectangle) with this class into the
query pane. Figure 1, center, shows the context menu available for nodes within the
visual diagram. The attributes, ascribed to the class (or to its superclass, or a subclass)
in the data schema, can be checked to be added to the query output via the Add Attrib-
utes context menu item (the attribute choice dialogue example is in Figure 1, right; the
(select this) option adds the class instance URI itself; the [+] button expands the choice
list to offer object property selection options, as well).
At this point the query SPARQL form can be generated or the query can be executed
over the provided SPARQL endpoint using the respective context menu commands.
1
http://viziquer.lumii.lv/schema-store/index.html
2
http://www.scholarlydata.org/sparql/
Figure 1. Query environment elements: clickable class tree (left), context menu (center) and
attribute addition dialogue example (right).
There are also separate context menu items for aggregation (e.g. count) option and
condition introduction into the query nodes. The full editing facilities of the visual
query element properties are available in the designated property pane that can be used
for structural element introduction, fine-tuning, editing and deleting alike.
There are two options of linked node introduction into a query: either by introducing
both nodes (typically, classes) into the query pane and then connecting the nodes by a
link from the diagram symbol palette, or the Add Link option from the class context
menu that suggests adding the links with schema-defined properties for the class, to-
gether with the property target classes. Both link creation options distinguish the class
join links from the nested query links (links with black bullets, as e.g. in Table 1 exam-
ples 4, 5, 6 and 7). The Add Link dialogue further on offers a wizard for fast introducing
of aggregation into the nested query, together with its result handling within the host
query, as e.g. in query 4 in Table 1.
The textual expression information entry within attribute expression and condition
fields and in line property paths is supported by a code completion option that is aware
of the expression focus placement with respect to the data schema and the field position
within the query. So, entering the property paths e.g. in queries 3, 5 and 6 in Table 1,
each property name in the path is gradually offered to the end user after the entry of the
previous property name, followed by the dot that indicates the continued navigation.
The query re-shape options involve a service for changing the query main class, and
a more fundamental Add Outer Query option that adds a non-data node ‘[ ]’ introducing
the outer query level around the current base query, as e.g. in query 7 in Table 1.
4 Discussion and Conclusions
An earlier user study with Computer Science master’s degree students has shown [9]
that for persons literate in computing the visual query composition is easier than corre-
sponding query writing directly in SPARQL. Still, the study indicated not very high
productivity in visual query creation (on average just 5.27 successful queries within 70
minutes [9]). To test the query composition environment improvements from [9] and
[11], we repeated the test (the visual notation part) with 28 similarly prepared partici-
pants (the students of the same study course the next year), and the average correctly
completed query score has risen to 7.5 under similar conditions. Meanwhile, the most
difficult query task with aggregating over attribute under the condition of existential
nested query that has reached only 25% (2 of 8) correct score among those attempted
the query in [9], did receive the rather unconvincing 45% (9 of 20) here. The analysis
of the study results has allowed to identify issues in query composition, resolved in the
current tool release, leading to a hope that the ViziQuer tool can ease the query com-
position task over RDF data stores at least for technically literate persons.
The ViziQuer tool is open source. To ease the local ViziQuer server usage, a pre-
built Docker environment with the tool is offered. The further usability improvements
including the integrated and more efficient data schema retrieval is work in progress.
References
1. 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.
Universal Access in the Information Society, March 2016, Volume 15, Issue 1, pp 129–152.
2. Zviedris, M., Barzdins, G.: ViziQuer: A Tool to Explore and Query SPARQL Endpoints. In:
The Semantic Web: Research and Applications, LNCS, Volume 6644, pp. 441-445, (2011)
3. B.Kapourani, E. Fotopoulou, D. Papaspyros, A. Zafeiropoulos, S. Mouzakitis, S.Koussou-
ris., Propelling SMEs Business Intelligence Through Linked Data Production and Consump-
tion, In OTM OTM 2015 Workshops pp 107-116.
4. Haag, F., Lohmann, S., Siek, S., Ertl, T.: QueryVOWL: Visual Composition of SPARQL
Queries. In: The Semantic Web: ESWC 2015 Satellite Events. LNCS, Vol.9341, pp. 62-66.
Springer, (2015), http://vowl.visualdataweb.org/queryvowl/
5. K.Cerans, J.Barzdins, A.Sostaks, J.Ovcinnikova, L.Lace, M.Grasmanis and A.Sprogis. Ex-
tended UML Class Diagram Constructs for Visual SPARQL Queries in ViziQuer/web In
Voila!2017, CEUR Workshop Proceedings, Vol.1947, (2017) pp.87-98.
6. G. Vega-Gorgojo, M. Giese, S. Heggestoyl, A. Soylu, A. Waaler. PepeSearch: Semantic
Data for the Masses. In: PLoS ONE 11(3): e0151573. doi: 10.1371/journal.pone.0151573,
2016. http://dx.doi.org/10.1371/journal.pone.0151573
7. A.Khalili, A.Meroño-Peñuela. WYSIWYQ --- What You See Is What You Query. In
Voila!2017, CEUR, Vol.1947, (2017) pp.123-130. http://ceur-ws.org/Vol-1947/paper11.pdf
8. S.Ferré: Sparklis: An expressive query builder for SPARQL endpoints with guidance in nat-
ural language, Semantic Web, 2017, Vol 8, pp 405-418
9. Čerāns, K., Šostaks, A., Bojārs, U., Bārzdiņš, J., Ovčiņņikova, J., Lāce, L., Grasmanis, M.
and Sproģis, A., ViziQuer: A Visual Notation for RDF Data Analysis Queries. In Research
Conference on Metadata and Semantics Research. Springer CCIS, Vol.846, pp.50-62, 2018
10. SPARQL 1.1 Query Language. W3C Recommendation 21 March 2013,
http://www.w3.org/TR/2013/REC-sparql11-query-20130321/
11. Čerāns, K., Šostaks, A., Bojārs, U., Ovčiņņikova, J., Lāce, L., Grasmanis, M. Romāne, A.,
Sproģis, A., Bārzdiņš, J. ViziQuer: A Web-Based Tool for Visual Diagrammatic Queries
Over RDF Data. In: ESWC 2018 Satellite Events. LNCS, vol 11155, pp. 158-163, 2018.
12. A. L. Gentile and A. G. Nuzzolese. cLODg - Conference Linked Open Data Generator. In
ISWC 2015 Posters & Demonstrations Track, CEUR-WS.org, Vol. 1486, 2015.