Please use this identifier to cite or link to this item:
Title: A constructive theory of automated ideation
Issue Date: Apr-2014
Publisher: Imperial College London;Computing
Description: In this thesis we explore the field of automated artefact generation in computational creativity with the aim of proposing methods of generation of ideas with cultural value. We focus on two kinds of ideas: fictional concepts and socially embedded concepts. For fictional concepts, we introduce a novel method based on the non-existence-conjectures made by the HR automated theory formation system. We further introduce the notion of typicality of an example with respect to a concept into HR. This leads to methods for ordering fictional concepts with respect to three measurements: novelty, vagueness and stimulation. We ran an experiment to produce thousands of definitions of fictional animals and then compared the software's evaluations of the non-fictional concepts with those obtained through a survey consulting sixty people. The results showed that two of the three measurements have a correlation with human notions.For socially embedded concepts, we apply a typicality-based classification method, the Rational Model of Classification (RMC), to a set of data obtained from Twitter. The aim being the creation of a set of concepts that naturally associate to an initial topic. We applied the RMC to four sets of tweets, each corresponding to one of four initial topics. The result was a set of clusters per topic, each cluster having a definition consisting of a set of words that appeared recurrently in the tweets. A survey was used to ask people to guess the topic given a set of definitions and to rate the artistic relevance of these definitions. The results showed both high association percentage and high relevance scores. A second survey was used to compare the rankings on the social impact of each of the definitions. The results obtained show a weak positive correlation between the two rankings. Our experiments show that it is possible to automatically generate ideas with the purpose of using them for artefact generation. This is an important step for the automation of computational creativity because most of the available artefact generation systems do not explicitly undertake idea generation. Moreover, our experiments introduce new ways of using the notion of typicality and show how these uses can be integrated in both the generation and evaluation of ideas.;Open Access
Other Identifiers:
Type Of Material: OTHER
Appears in Collections:Faculty of Engineering

Files in This Item:
Click on the URI links for accessing contents.

Items in HannanDL are protected by copyright, with all rights reserved, unless otherwise indicated.