Little Search Game: Lightweight Domain Modeling - Semantic Acquisition Games: Harnessing Manpower for Creating Semantics - page 50

Fig. 4.1 Screenshot capturing the interface of the Little search game. On the left pane, player enters
the negative terms. He may observe his relative progress on the chart in the right
search terms to it. The negative search terms (often declared in search interfaces by
minus sing prefix) cause the search engine to remove all results from the query yield,
that contain the particular negative search term. The player's goal, is to come up
with such negative search terms (to the given task term) that will cause a maximum
possible decrease in number of results yielded by the extended query (according to
this, the player is scored). To achieve this, it is suitable for the player to use search
terms semantically related to the query term, as these often co-occur together on
the Web. The game then constructs a term relationship network by mining the game
query logs.
The typical gameplay scenario of the LSG works as follows: (refer also to the
game screenshot Fig. 4.1 ):
1. On game start-up, a task term is selected for the player, for example “star”. The
number of web search results is displayed for this term (it is obtained by real
query). For this particular term, it is nearly 500millions of results.
2. The player then comes up with the negative search terms, entering them into
prepared text fields. For example, the player decides to use word “movie” as
negative term, creating query “star—movie”, because he considers these two
words to be semantically related and are thus co-occurring often on the web. As
a result of entering the “movie” word, the number of results drops by 100millions