Game Development Reference
In-Depth Information
are not so obviously related to the given word. Authors claim relatively large number
of explored relationships (800 thousand raw data instances).
Vickrey et al. also made an interesting analysis of the types of the word-to-word
relationships they retrieved from the Free Association (which, as we will see, is a very
similar issue to one we solved in our own SAG research). Using a predefined set of
possible relationships (e.g., “has”, “made of” or “opposite of”), they manually typed
500 of their relationships. The most numerous were miscellaneous relationships
(20%), but synonyms and hypernyms ranked high too. An interesting option was the
possible linkage of the two words with two simple relationships, through some third
word (3%). Unfortunately, authors have not analyzed the possibilities of automated
relationship type mining.
3.4.6 Akinator
The
Akinator
3
is a unique SAG for populating the knowledge base in the domain
of popular or otherwise widely known persons. It is a single player game in which
the artificial social agent, the
Akinator
, asks the player to secretly think of some
relatively famous personage. Then, he asks the player a series of yes-or-no questions
(with possible answer of
unsure-yes
,
unsure-no
and
unsure
, see Fig.
3.5
).
Based on player's answers, the game filters it's existing database of persons and
tries to select one that best fits the criteria given by player. Then, the game offers
the player its guess (which is in many cases a correct answer—a feature that greatly
attracts the players). If the player confirms the correct answer, the agent wins, other-
wise, the game continues. After three wrong guesses of the social agent, the player
wins. In that case, the Akinator asks the player to type in name of the person he has
in mind (see Fig.
3.6
). The player may also suggest a new question the Akinator
should ask the players. It is also apparent that game's knowledge base and reasoning
work on the probabilistic principle and is tolerant to player errors (wrong answers).
If, for example, the player is having a American president in mind and answers the
question “Did your character came from America” with “no”, the game is able to
overcome it after asking several other questions, effectively invalidating the wrong
answer. From the semantics creation and the SAG design point of view, the Akinator
represents a successful case of social agent backed with an existing knowledge base.
3.5 Discussion
In the last decade, we have observed proliferation of semantics acquisition games.
SAGs represent a dynamic field, which firmly established itself within the crowd-
sourcing research domain and the problem research domains with which the SAGs
Search WWH ::
Custom Search