Game Development Reference
In-Depth Information
about time and order the combination was suggested in the game, neither we consider
the number of search results yielded in the game. The only relevant information is
that at some point, the player considered the two terms related.
After we collect the votes, we fold them according to term combinations they
represent. If a particular term-term combination is supported by more players than
specified by a threshold (a process parameter, set to 5 in our experiments), an oriented
relationship is created, with task term as a source term and negative search term
as target term. Moreover, each such term relationship is decorated by a ratio of
two numbers, which together constitute a relative “strength” or “weight” of the
relationship: the number of votes that relationship has received (
ω
p
) and the total
number of votes that share the same source term as this relationship (
ω
t
).
The term network of the
Little Search Game
is then a graph
G
having the set of
nodes
V
(representing terms) and set of edges
E
:
G
(
V
,
E
)
V
:
T
∪
N
E
:
V
×
V
× N × N
The edge represents a term relationship
e
;
e
∈
E
.
e
is a quartet
e
=
(
t
,
n
,ω
t
,ω
p
)
where
t
ω
p
represents the relative weight
w
of the
edge to other edges outgoing from the same node:
∈
T
and
n
∈
N
. The ratio
ω
t
:
=
ω
p
ω
t
w
4.3 Little Search Game Evaluation
4.3.1 Game Deployment
The game was deployed as a browser implementation, called Little Google Game,
as it used the Google search engine to commence the game queries (yet any web
search engine is applicable). The Google was our first choice (for it had broadest
web space covered), but later we switched to we switched the search engine to
Bing service (offered by Microsoft). We did this, paradoxically, due to the increased
sophistication of the Google search, which no longer followed the Boolean logic of
negative search clauses. The Bing service, on the other hand offered better results.
We deployed the game for several times. First, the game was played with an initial
set of 20 arbitrary chosen task words (nouns of the common knowledge domain, e.g.
“water”, “castle”, “brain”). Up to 300 games with 2,000 submitted queries were
played by 30 players. The number of recorded logs was therefore relatively small,
Search WWH ::
Custom Search