Computational Creativity Conceptualisation Grounded on ICCC Papers Senja Pollak 1 , Biljana Mileva Boshkoska 1,3 , Dragana Miljkovic 1 , Geraint A. Wiggins 2 , Nada Lavraˇ c 1,4 1 Dept. of Knowledge Technology, Joˇ zef Stefan Institute, Ljubljana, Slovenia 2 Computational Creativity Lab, Queen Mary University of London, London E1 4NS, UK 3 Faculty of Information Studies, Novo Mesto, Slovenia 4 University of Nova Gorica, Nova Gorica, Slovenia {senja.pollak,biljana.mileva,dragana.miljkovic,nada.lavrac}@ijs.si; geraint.wiggins@qmul.ac.uk Abstract In information science, it is considered that domain con- ceptualization can be realized by (one or several) ontolo- gies. This paper presents a method of semi-automated do- main conceptualization, where the domain of interest is Com- putational Creativity (CC). Grounded on papers, which were published in six consecutive years since 2010 in the Proceed- ings of International Conferences on Computational Creativ- ity (ICCC), this paper proposes a tentative conceptualization of the CC domain. Some additional properties of the CC do- main are studied, analysed by means of fully mechanical or semi-automated information extraction and dependency anal- ysis techniques. This approach affords an interesting oppor- tunity for automated historiography of a research field. Introduction As a sub-field of Artificial Intelligence research, Com- putational Creativity (CC) is concerned with engineering software that exhibits behaviours which would reasonably be deemed creative (Wiggins, 2006; Colton and Wiggins, 2012). A part of CC research addresses Concept Creation Technology, concerned with engineering software that ex- hibits creative conceptualization behaviour. For a given domain, whose conceptual space (Boden, 2004) is closed, pre-defined and yet unexplored, it is inter- esting to study computational means for automated (or semi- automated) domain conceptualization. In the current re- search, we use the term conceptualization in alignment with its standard use in information science: a conceptualization is defined as “an abstract (simplified) view of some selected part of the world, containing the objects, concepts, and other entities that are presumed of interest for some particular pur- pose and the relationships between them.” Domain concep- tualization is, in information science, frequently realized by manually defining (one or several) ontologies formally de- scribing the domain of interest (Gruber, 1993; Smith, 2003). Manual construction of ontologies represents a significant investment of human resources when used for modelling a new domain. Therefore, methods for (semi-)automated ex- traction of domain knowledge from unstructured texts were developed, including automated taxonomy construction de- scribed by Velardi, Faralli, and Navigli (2013). While an ontology is a “formal, explicit specification of a shared conceptualization” (Gruber, 1993), represented as a set of domain concepts and the relationships between them, a so-called topic ontology is a set of domain top- ics or concepts—formed of related documents—represented by the most characteristic topic keywords and related by the subconcept-of relationship (Fortuna, Grobelnik, and Mladeni´ c, 2007). The task addressed in this paper is semi- automated construction of a topic ontology from documents in the area of computational creativity. CC domain conceptualization has not been substantially addressed in the CC literature. Jordanous and Keller (2012) used automated natural language processing methods and a statistical measure of association to identify words related to creativity (in general, not specifically CC). They clustered the words into semantically-related groups by using a lexical similarity measure, resulting in an ontology of creativity 1 ). Others presented extraction of creativity concepts related to, e.g., sub-fields of creativity (Agres et al., 2015) and creativ- ity evaluation (van der Velde et al., 2015). This approach to the study of collections of documents opens the prospect of an automated historiography of the field of computational creativity, an idea which constitutes a satisfyingly recursive application of the research outputs of that area of interest; a similar exercise has been under- taken within the Association for Computational Linguistics (Anderson, McFarland, and Jurafsky, 2012). Here, we illus- trate, with real computational examples, the kinds of analy- sis (e.g., diachronic comparisons of conceptualisation) that would be used for such studies. The current paper presents a method of semi-automated domain conceptualization, where the domain of interest is Computational Creativity (CC). The paper proposes a con- ceptualization of CC grounded in papers published in the Proceedings of International Conferences on Computational Creativity (ICCC). Some additional properties of the CC do- main are studied, obtained by means of information extrac- tion and dependency analysis techniques. The experimental data is presented, followed by the results of CC domain con- ceptualization and time dependency analysis. The Computational Creativity Domain This section describes the data used in the experiments, to- gether with initial domain understanding achieved through automated terminology extraction. 1 http://purl.org/creativity/ontology 128 123 Proceedings of the Seventh International Conference on Computational Creativity, June 2016