=Paper=
{{Paper
|id=Vol-1963/paper474
|storemode=property
|title=Clover Quiz: a Mobile Trivia Game Based on DBpedia Data
|pdfUrl=https://ceur-ws.org/Vol-1963/paper474.pdf
|volume=Vol-1963
|authors=Guillermo Vega-Gorgojo
|dblpUrl=https://dblp.org/rec/conf/semweb/Vega-Gorgojo17
}}
==Clover Quiz: a Mobile Trivia Game Based on DBpedia Data==
https://ceur-ws.org/Vol-1963/paper474.pdf
Clover Quiz: a Mobile Trivia Game Based on
DBpedia Data
Guillermo Vega-Gorgojo1,2
1
Don Naipe
2
Department of Informatics, University of Oslo
Abstract. This demonstration presents Clover Quiz, a turn-based mul-
tiplayer trivia game for Android devices with more than 200K multiple
choice questions (in English and Spanish) about different domains gen-
erated out of DBpedia. Instead of live queries, questions are created off-
line through a data extraction pipeline and a versatile template-based
mechanism. A back-end server manages the question set and the asso-
ciated images, while a mobile app has been developed and released in
Google Play on March 2017 under the names ‘Clover Quiz’ in English and
‘Trebial’ in Spanish. The game is available free of charge and has been
downloaded by more than 5K users at the time of this writing. During
this period, more than 609K questions have been posed, and the overall
rating of the game is 4.3 out of 5.0. Clover Quiz thus demonstrates the
advantages of semantic technologies for collecting data and automating
the generation of multiple choice questions in a scalable way.
1 Introduction
DBpedia is a large-scale and multilingual knowledge base that constitutes the
main hub of the Semantic Web. An appealing application case is the generation
of questions for trivia games from DBpedia. Some preliminary attempts can be
found in the literature [1,2,3], but these initiatives have fallen short due to simple
question generation schemes that are not able to produce varied, large, and
entertaining questions. Specifically, supported question types are rather limited,
reported sizes of the generated question sets are relative low (in the range of
thousands), and no user base seems to exist. Moreover, some of these works
create the questions by submitting live queries to the public DBpedia endpoint,
hence latency is too high for an interactive trivia game, as reported in [2].
The hypothesis is that creating questions from DBpedia can be significantly
improved by splitting this process in a data extraction and a versatile question
generation stages. This approach can produce varied, numerous, and high-quality
questions by declaratively specifying the classes and question templates of the
domains of interest. The generated questions can then be hosted in a back-end
server that meets the latency requirements of an interactive trivia game. The
target case is a turn-based multiplayer trivia game in which two players compete
over a clover-shaped board by answering multiple choice questions from different
domains. This demonstration presents such a game, named Clover Quiz, that is
available for Android devices through Google Play.
2
2 Data Extraction and Question Generation
Since the game is purposed for mobile devices, Clover Quiz employs multiple
choice questions with 4 options to limit user typing. To generate the questions,
the data of interest is gathered from DBpedia. In a first stage, a domain specifi-
cation file is authored with the classes in the domain of interest, e.g. Museum, or
Painting in Arts. Each class is associated with a SPARQL query for retrieving
the corresponding entities. The specification file also identifies the literals to be
extracted for the entities of a target class –like labels, years, or image URLs–
as well as relations between entities of different classes. Next, a script takes a
specification file as input and systematically queries DBpedia to retrieve the
data available of the target domain. Specifically, this script gathers the entities
belonging to each class, their literals, and their relations with other entities. The
output of this stage is a file with a JSON object for every entity found.
The data extraction pipeline includes a category annotation stage based on
the Wikipedia categories extracted for each entity. Wikipedia contributors an-
notate articles with suitable categories that are organised into a hierarchy that
reflects the notion of “being a subcategory of”. Since there are more than one
million Wikipedia categories, a script automatically gathers a relevant subset
of the domain of interest by posing SPARQL queries to retrieve the annotated
categories of the extracted entities and then the expanded category set. Fi-
nally, another script takes as input the extracted data, the expanded categories,
and the Wikipedia categories of interest, e.g. dbc:Baroque_paintings for the
Painting class. For every candidate category, the script obtains the expanded
list of subcategories and annotates the appropriate entities of the target class.
With the obtained annotated data files, questions are then generated using
multilingual question templates. A template can involve either entities of a single
class or of two classes, depending on the template type. These are the follow-
ing: image, e.g. Which is the painting of the image?; boolean, e.g. Which is the
modernist building?; boolean negative, e.g. Which is NOT an Ancient Greek
sculptor?; group, e.g. Which is the artistic style of the painter {{painter.label}}?
(options: Gothic, Renaissance, Baroque, Mannerist, Romantic); date, e.g. When
was {{painter.label}} born?; greatest, e.g. Which country has the largest pop-
ulation?; numeric, e.g. Which is the population of {{city.label}}?; relation,
e.g. Who is the painter of “{{painting.label}}”?; and relation negative, e.g.
Which castle is NOT in {{country.label}}?.
A script is in charge of generating the questions by evaluating every template.
Briefly, the script first obtains the set of entities that comply with the template
requirements, e.g. paintings with animals from the Baroque period. It will then
generate a question for each occurrence by getting the image URL (to support
the question) and the label of the painting (this will be the correct answer).
Finally, a list of distractors is prepared from the target set. Fig. 1(top) shows two
sample questions generated through this process. Although not described here
due to space constraints, a question difficulty estimator is employed in order to
adjust the difficulty of the questions to the player skills. Table 1 presents some
aggregated figures of the question set obtained for the target domains.
3
Table 1: Summary of the question generation process for the different domains.
Animals Arts Books Cinema Geo Music Tech TOTAL
# of templates 125 269 387 295 724 767 374 2,941
# of questions (Spanish) 15,342 18,121 23,580 49,208 24,086 36,136 21,014 187,487
# of questions (English) 15,347 27,523 46,403 50,199 24,484 36,075 21,017 221,048
3 Demonstration Overview
The obtained question set was deployed in a MongoDB database. Node.js was
employed to handle question requests, while an Nginx server was configured
to host the game images (37K low-resolution images, totalling 1.12 GB) and
to forward question requests to Node.js. As for the mobile app, I coded an
Android version of the game that can be played in phones and tablets –see sample
snapshots in Fig. 1. The mobile app supports two-player matches, single-player
mode, statistics, leaderboards and achievements.
The game was released for Android devices on March 11, 2017 under the
names ‘Clover Quiz’ in English and ‘Trebial’ in Spanish. It is available free of
charge through Google Play at https://play.google.com/store/apps/details?
id=donnaipe.trebial and is part of the catalogue of Don Naipe, a sole propri-
etorship company specialized in Spanish card games for mobile devices. Clover
Quiz was promoted with an in-house ad campaign that ran from March 13 to
March 16, i.e. other Android games by Don Naipe1 showed interstitial ads about
Clover Quiz. At the time of this writing (July 2017), more than 5K users have
downloaded the game, requesting more than 609K questions. Clover Quiz users
have also given feedback through Google Play. Specifically, the average rating is
4.3 out of 5.0 and users’ comments are generally very supportive.
Acknowledgements
This work has been partially funded by the Norwegian Research Council through
the SIRIUS innovation center (NFR 237898).
References
1. Bratsas, C., Chrysou, D.E., Eftychiadou, E., Kontokostas, D., Bamidis, P., Anto-
niou, I.: Semantic Web game based learning: An i18n approach with Greek DBpedia.
In: Proceedings of the 2nd LiLe Workshop. Lyon, France (Apr 2012)
2. Mütsch, F.: Auto-generated trivia questions based on DBpedia data (Feb 2017),
URL: https://github.com/n1try/linkeddata-trivia, last accessed July 2017
3. Waitelonis, J., Ludwig, N., Knuth, M., Sack, H.: Whoknows? Evaluating linked data
heuristics with a quiz that cleans up DBpedia. Interactive Technology and Smart
Education 8(4), 236–248 (2011)
1
https://play.google.com/store/apps/developer?id=Don+Naipe
4
Fig. 1: Sample questions from the Arts and Animals domains (top). Sample snap-
shots of the main and game screens of Clover Quiz (bottom).