Story Generation Based on Change of Social Relations by Selecting Event Fragments from Human-Outsourcing Stories

• Main Authors: Chu, Li-Yang, 褚立陽
• Other Authors: 蘇豐文
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/04198891595251236465

碩士 === 國立清華大學 === 資訊工程學系 === 102 === The story plots without considerations about social friendship may be not reasonable or its feasibility may be limited. I propose a friendship resolution that can provide a plot that which accomplishes the required friendship change. I collect stories from human outsourcing, and transit the stories into a designed presentation which could be processed by program. Then I slice the stories according to the inner event’s causality or co-meaning relation and extract story fragments. The “friendship impact” and “friendship level to appear” are estimated. Finally, we can generate the plot by put fragments in appropriate friendship level and estimate the current friendship with the friendship impact. The experiment which ask form fillers to sort the rationality between a human-written, a model generated and a random generated stories has shown that the model generated story was even better than human-written ones. And this model is generic to handle requires in open domain.