Teaching machines to understand human com- munication is one of the central goals of arti- ficial intelligence. Psychological research in- dicates that human associations are an essen- tial requirement to understand human commu- nication. In this paper the hypothesis is pre- sented that simulating human associations with the help of Linked Data could improve text un- derstanding capabilities of machines. To inves- tigate whether human associations can be sim- ulated with Linked Data, two preliminary prob- lems are identified: (i) A reasonable ground truth for human associations is lacking and (ii) hu- man associations have different strengths while Linked Data treats all triples equally and does not provide edge weights. To overcome these prob- lems, two ideas for web games in accordance with Luis von Ahn’s Games with a Purpose are proposed trying to turn the tedious acquisition processes into fun games. The resulting datasets are then to be used for quantitative comparisons of human associations and Linked Data.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below