Introduction - Semantic Acquisition Games: Harnessing Manpower for Creating Semantics - page 2

to come up with good-enough search queries to satisfy their needs (usually, this
happens when users are not familiar with the domain they are searching). In such sit-
uations, the Semantic Web would allow to create easy-to-browse information space
abstractions that help the searchers to orient themselves within the domain they are
not familiar with.
Unfortunately the vision is far from the fulfillment. The Semantic Web does not
exists in the scale needed to become a background for a prevailing search paradigm
(although we can observe promising initiatives such as linked data creation [ 1 ]). The
reason is that creation of proper web resource metadata requires human expertise and
extensive work (which may go costly in web scale) or sophisticated automated meth-
ods doing the same job (which have presently only limited capabilities). Extensive
effort is also needed to create the domain models (to cover the layer from the top)
either by manual means or in devising automated methods. This makes the seman-
tics acquisition an attractive field with a plethora of existing approaches and research
opportunities.
We roughly split the existing approaches to semantics acquisition into the follow-
ing categories (although they also tend to overlap and complement each other):
1. Expert ( human computation ) work. Comprises work of domain experts, who
create either annotations of resources or domain ontologies (e.g., project Cyc
[ 8 ]). They may also include other approaches where metadata are created with
expertise of a single individual. Manual semantics creation delivers high quality
results, but cannot cover the vastness of the Web without being too expensive.
2. Crowd ( human computation ) work—the crowdsourcing is still human-originating
semantics creation, but capable of delivering semantics in high quantity, although
with quality varying in terms of generality (they do not work well in specialized
domains). The “crowd” means that there are many knowledge-contributing indi-
viduals in the process. This is usually possible thanks to the fact that crowd
members contribute only as a by-product of other primary activity they are
motivated to do (e.g., contributing image annotations while organizing their
image galleries). The second reason is that crowd members are “non-experts”
regarding the (semantics acquisition) job they perform. To keep quality outputs
in these conditions, several validation mechanisms are being used, for exam-
ple multiple user agreement [ 2 , 13 ]. The general crowdsourcing also includes
game-based approaches, i.e. crowdsourcing games (sometimes called games with
a purpose [ 20 ])—specially designed games that transform work-like tasks to
entertaining experiences.Whenwe use crowdsourcing games for semantics acqui-
sition, we call them semantics acquisition games (SAGs). The field of SAGs is
the primary field of interest of this work.
3. Machine ( automated ) approaches for semantics acquisition implement various
natural language processing techniques, data mining and machine learning in
order to annotate resources or extract domain knowledge [ 6 , 12 , 14 , 16 , 21 ].
While capable of delivering even web scale quantity of information, they often
suffer from inaccuracies, mainly due to the heterogeneous nature of the Web and
natural language, which they cannot effectively sustain. Nevertheless, they are