Summary: | Interactive storytelling systems integrate artificial intelligence methods to generate narratives for the user (reader or player) to experience a story line that is not completely predetermined by the author. One method for doing this is planning where, in the context of interactive storytelling, the method is given an initial and goal state of the story world. It then generates the narrative by searching through available story actions to find a path to the authored goal state, resulting in a desired story experience. Traditionally, data representations for these planning algorithms present restrictions that limit the expressive power of the author. The SAGA Project at Reykjavík University is an example of an interactive storytelling system where such a data representation is used. In this thesis the SAGA project is modified to use a new data representation to alleviate some of these restrictions. This was done in two steps. First, the new data representation and planning system were integrated into the system to replicate the output of the existing implementation. Second, the new data representation was extended to be able to handle a more flexible way of expression. Evaluation shows that the new approach is successful in providing more expressive power to the author. However, performance tests suggest that the new approach is less efficient than the previous implementation. Kerfi fyrir gagnvirkar frásagnir innleiða gervigreindaraðferðir til þess að búa til sögur svo að notandi (lesandi eða spilari) getur upplifað söguþráð sem höfundur hefur ekki algjörlega ákveðið fyrirfram. Ein aðferð til að gera þetta er áætlanagerð sem, þegar kemur að gagnvirkum frásögnum, er gefin upphafspunktur og endapunktur söguheimsins. Aðferðin býr síðan til söguna með því að leita í gegnum aðgerðir sem í boði eru til að finna leið að endapunkti höfundar og til verður saga sem höfundur sóttist eftir. Þessar aðferðir áætlanagerða nota yfir\-leitt hefðbundna framsetningu á gögnum sem hafa í för með sér takmarkanir á því hvernig ...
|