通讯作者:
版权声明: 2020 【-逻*辑*与-】#x000a9; 2020 Totem Publisher, Inc. All rights reserved.
展开
Abstract
Creativity can be challenging in the idea generation process as it is hard to formalise and control. However, there are requirements for creative ideas in many fields. Generating new and inspirational ideas is mainly manual work, and it usually happens in the individual human mind or a group of persons. There is a lack of software systems to generate creative ideas automatically. In this paper, a prototype software system based on a semantic reasoning method is proposed for assisting creativity in the general idea generation process. The kernel algorithm of the system is a set of inference rules designed on the basis of semantic computing technologies and creativity techniques, which is the core of the semantic reasoning. The fundamental information supporting the inference are domains knowledge managed as ontology bases. Furthermore, a major recommender is designed and implemented by employing the proposed idea creation method to enhance the inspiration level of university choice for teenagers. As a prototype software system, the developed major recommender application proves the feasibility and innovation of the proposed method.
Keywords:
© 2020 Totem Publisher, Inc. All rights reserved.
There has been an ongoing interest in creating new ideas for decades, as creative ideas are the basis of all innovations. Thus, it is obvious to see how important new ideas and the techniques of generating them are in our lives. Idea generation, also called ideation, is generally considered as the creative process of generating, developing, and communicating new ideas, while the ideas can be represented in many forms including drawn pictures, verbal language, rhythms and melodies, and written text. Many efforts have been done to study and promote idea generation methods and techniques for human beings. However, there have only been sporadic attempts to study the creativity aspect of idea generation from computer technology's perspective, and many of them focus on psychological creativity (P-creativity) but not historical creativity (H-creativity).
To distinguish it from general idea generation, idea creation as used in this paper specifically refers to computer-ideation. The goal of this research is to propose a set of inference rules for assisting and enhancing creativity in the idea creation process, especially for H-creativity. The inference rules are designed based on semantic computing technologies and creative elements. Specifically, ontology base is adopted as the representation of knowledge as well as main idea creation activities, and atomic operations are designed based on creativity techniques. The reasoning outcomes are ideas balanced at valuable and surprising. A prototype of the creative major recommendation system is developed based on the proposed semantic reasoning method to assist students' decision making when choosing their university major while considering their personal information. In the prototype, the outcomes of a set of input data are two conventional majors and two inspirational majors. The former is the suggestion that can be seen on other major recommenders, while the latter is the most likely unexpected result; these inspirational recommendations distinguish the system from the others. The knowledge-based semantic reasoning method enhances the capability of expanding other knowledge related to major choosing and considers creative elements as kernel factors. The results are evaluated from the perspectives of novelty, usefulness, and surprising. Furthermore, the overall creativity degrees are evaluated according to adjusted weights and calculated degrees of novelty, usefulness, and surprising. The evaluation illustrates that the proposed knowledge-based semantic reasoning method works effectively at enhancing an idea's creativity.
2.1. Creativity, Creativity Techniques, and Idea Generation
Creativity is an extremely important facet of life and is considered regardful on new ideas. It is believed that creativity is a highly complex process that can be considered as the ultimate human activity. Therefore, creativity is difficult to formalise and control [1]. It has been argued that creativity is a human prerogative and creative ideas can only be generated in the human mind [2-4]. Boden defined that an idea can be called new or creative from two perspectives, the objective view and the subjective view, which are derived from two kinds of creativity: H-creativity (short for historical creativity) and P-creativity (short for psychological creativity) [5-6]. H-creativity is fundamentally novel with respect to the whole of human history, and P-creativity is the personal kind of creativity that is novel in respect to the individual mind [7-8]. According to Boden's theory, human beings are good at generating new ideas from the subjective perspective. However, it is hard to achieve H-creativity, as humans are unable to gain and archive the vast and fast-changing domain knowledge on time in this era of information explosion.
Therefore, with the computer's strong calculation capability, it is reasonable to assist human creativity by using a carefully designed idea creation system. In the idea generation related studies, there are efforts that either implicitly or explicitly assure Osborn's conjecture that if people generate more ideas, then they will produce more good ideas [9]. However, other researchers have argued that there are no relationships between idea quality and idea quantity [10]. Apart from the research work focusing on idea quantity and quality, there has also been work contributing to the dialectical process of knowledge creation or individual attribution. Furthermore, in the last two decades, there have been many studies for system-assisted team idea generation using brainstorming and related techniques [11]. Overall, most works consider a good idea as one that is feasible to implement, that would attain a goal, and that would not create new unacceptable conditions. Thus, they have neglected creativity's importance to form a novel idea. There have been few attempts applying creativity techniques to a variety of approaches for supporting a group idea generation process [12-13].
At a higher level of abstraction, creativity techniques are divided into three categories: exploratory creativity, transformational creativity, and combinational creativity [14]. Exploratory creativity represents an existing conceptual space conducting the research to achieve creativity, transformational creativity involves creating a new conceptual space based on transforming an existing conceptual space, and combinational creativity aims at combining similar ideas and thoughts to establish new methods [15]. Specifically, there are many concrete creativity techniques in the three categories. Smith's article [16] summarised and analysed 172 creativity techniques as human idea generation methods and identified active ingredients based on the techniques. Because its analysis covered most of the famous and general creativity techniques for idea generation, and the abstracted active ingredients are well-accepted in the area of human idea generation, it is worth being employed in this research as the basic information of creativity techniques.
Studies on the creativity techniques are mostly either for idea generation happening in the human mind, team discussion [13, 16-18], or other research areas (e.g., product design), but not for idea creation [19-22]. Only a few research studies have made efforts in applying creativity techniques for idea generation into computer systems [23- 24]. However, this kind of research study on creativity techniques generally involves sporadic attempts and a lack of detail. This research proposes a set of creativity elements based on analysing creativity techniques and designs semantic inference rules supported by ontology bases and the creativity elements.
2.2. Semantic Computing Techniques
Semantic computing addresses the derivation and matching of the semantics of computational content to that of naturally expressed user intentions in order to retrieve, manage, manipulate, or even create content, where "content" may be anything including videos, audio, text, software, hardware, networks, and processes [25]. Semantic computing technologies refer to many kinds of technology (e.g., artificial intelligence, natural language processing, software engineering, data and knowledge engineering, semantic web, etc.) to facilitate activities of computing content based on machine-processable semantics ("meaning", "intention") [25-26].
Among the various technologies, description logic is a fundamental technique for knowledge representation and deductive inference, which is adopted in this research to support domain knowledge base and semantic reasoning. Description logic is a family of knowledge representation formalisms that represent the knowledge of an application domain by first defining the relevant concepts of the domain and then using these concepts to specify properties of objects and individuals occurring in the domain [27]. As the name indicates, one of the characteristics of these languages is that, unlike some of their predecessors, they are equipped with formal, logic-based semantics. Another distinguishing feature is the emphasis on reasoning as a central service: reasoning allows one to infer implicitly represented knowledge from the knowledge that is explicitly contained in the knowledge base. Description logic supports inference patterns that occur in many applications of intelligent information processing systems, and that are also used by humans to structure and understand the world through the classification of concepts and individuals. Classification of concepts determines sub-concept/super-concept relationships (called subsumption relationships in description logic) between the concepts of a given terminology, and thus allows one to structure the terminology in the form of a subsumption hierarchy. This hierarchy provides useful information about the connection between different concepts, and it can be used to speed up other inference services. The classification of individuals (or objects) determines whether a given individual is always an instance of a certain concept (e.g., whether this instance relationship is implied by the description of the individual and the definition of the concept). It thus provides useful information on the properties of an individual. Moreover, instance relationships may trigger the application of rules that insert additional facts into the knowledge base. In short, description logic provides reasoning services for retrieving relations between concepts and individuals.
Because description logics are knowledge representation formalism, and since in knowledge representation one usually assumes that a knowledge representation system should always answer the queries of a user in a reasonable time, the reasoning procedures description that logic researchers are interested in are decision procedures; these procedures should always terminate, both for positive and for negative answers. Since the guarantee of an answer in finite time need not imply that the answer is given in reasonable time, investigating the computational complexity of a given description logic with decidable inference problems is an important issue. The decidability and complexity of the inference problems depend on the expressive power of the description logic at hand. On the one hand, very expressive description logics are likely to have inference problems of high complexity, or they may even be undecidable. On the other hand, very weak description logics (with efficient reasoning procedures) may not be sufficiently expressive to represent the important concepts of a given application. As mentioned in the previous chapter, investigating this trade-off between the expressivity of description logics and the complexity of their reasoning problems has been one of the most important issues in description logic research.
Description logics are descended from so-called structured inheritance networks [28-29], which are introduced to overcome the ambiguities of early semantic networks and frames. This line of research culminates in the development of
3.1. Framework
A knowledge-based idea creation framework consisting of three phases is proposed, as shown in Figure 1. The three phases are knowledge bases construction, semantic reasoning, and idea representation. The framework is for general idea creation to provide new, useful, and inspirational ideas in various domains.
The first phase is knowledge base construction, that is, mapping data into domain knowledge bases. Our research considering the structured text as source data, semi-structured text, and unstructured text is not considered presently because it requires natural language processing technology to achieve the mapping goal. This phase aims to gather and process data and to transform it into knowledge organised as OWL ontologies. As the source data is limited to structured text, the abstraction-based mapping method proposed in our previous paper [32] is applicable. At the end of this phase, domain knowledge bases are constructed as OWL files.
The second phase is semantic reasoning to infer idea elements from constructed knowledge bases. This phase is the main focus of this paper; the main contribution of this work is a set of inference rules, which is the kernel support of the semantic reasoning. Idea elements are the outputs of this phase that are represented as ontology concepts and relations. Section 3.2 shows the process of generating idea elements and indicates how the inference rules work in this phase. Considering previous discussions on related work, selected creativity techniques are employed in the rule designing process. Atomic operations are created on the basis of creativity techniques to guide rules' design, and specific inference rules are designed accordingly. Details of atomic operations and inference rules are depicted in Sections 3.3 and 3.4.
The third phase is idea representation, that is, formatting idea elements as the user-friendly outcome. As stated before, this method is for general idea creation. It can be applied in the various domains in different circumstances. Therefore, it does not need to specify a specific form or forms of an idea's expression. Due to the application's requirement, final ideas are represented in a suitable format such as tables, sentences, and so on. Because this research works on text, the created ideas generally are shown as a kind of text as well. Apart from the expression format, the evaluation, ranking, and election are activities covered in this phase. Due to the page limit, this paper focuses on the second phase of the framework, so details of this phase are omitted.
3.2. Idea Elements Generation
Idea element generation is the actual work in the semantic reasoning phase as it determines how the idea elements are produced and where the creativity features come from. Figure 2 is the detailed workflow of generating idea elements, which is the detailed process of Phase 2 in the proposed knowledge-based semantic reasoning framework (referring to Figure 1).
Reasoning explicit and implicit semantics in constructed knowledge bases realises convergent thinking and builds creative idea elements. It starts with input validation, which takes keywords entered by the user and the constructed domain knowledge bases as this process's input. Then, it checks keywords in the knowledge bases to judge whether they are valid inputs or not. If a keyword is a concept in one of the knowledge bases, it is considered as a valid input, which can be used in the following reasoning part directly. In contrast, if a keyword does not match any concept in the knowledge base, it goes to a process that derives the invalid keyword into concepts that exist in the knowledge bases and is relevant to itself. The concept inference requires support from a semantic dictionary, such as WordNet [33-34]. Then, after the user selects one of the inferred concepts manually, the selected concept is considered as a valid input for the following idea elements reasoning part.
The kernel activities of this phase are to "select inference rules" and "apply reasoning rules", which are grouped and marked as reasoning in the workflow. To gain creativity features for idea elements, designed inference rules are employed to support the reasoning part. Particularly, by looking at the valid input, domain knowledge bases, and the conditions on the inference rules, it selects suitable rules to apply. For example, if the valid input is one concept, the rules that require two concepts as parameters may not be applicable or additional activities are required to obtain more inputs. Then, the selected inference rules work one by one to generate many concepts and roles. The main purpose of the inference rules is to derive non-exists but hidden or potential knowledge from the domain knowledge bases, so that the results are not only relevant but also innovative and surprising. Eventually, the idea elements are concepts and roles inferred from the knowledge base.
3.3. Atomic Operation
Creativity is a crucial feature for new and innovative ideas; consequently, it needs to be considered in the semantic reasoning phase. Considering the review of studies and developments related to creative ideas and creativity, this research believes that creative ideas can be generated systematically; in other words, creativity can be generated by a system if there is a carefully designed method supporting it. In particular, this paper presents some atomic operations designed based on creativity techniques, and they are supposed to be specific actions coming from the creativity techniques to support the design of inference rules. The atomic operations carry creativity techniques into semantic reasoning to add novel, useful, and inspiration characteristics into computing idea elements.
According to the review of creativity techniques that are relatively associated with this research, the 172 creativity techniques summarised and analysed by Smith [16] are employed in this research as the foundation of atomic operations. Although all the 172 creativity techniques are related to idea generation, not all are suitable for our purpose because some are only suitable for human minds (e.g., "mental simulation" technique), and others require group discussion (e.g., "drawing room" technique). Hence, only some techniques are selected to support atomic operations. The selected techniques are generally applied, widely discussed, and considered more related to this research. Figure 3 shows how the selected creativity techniques are transformed into atomic operations.
Particularly, the selected techniques are categorised into three groups according to their application domains including "creativity techniques for human ideation", "creativity techniques for problem-solving", and "creativity techniques for design", which are considered as the basis of atomic operations. The middle of Figure 3 contains two kinds of analysis, decomposition and classification, so that the creativity techniques can be broken down or classified into operations that are basic enough as indecomposable and disjoint operations.
Some of the selected creativity techniques are broken down into basal operations since they are abstract techniques; parts of the repetitive techniques are classified into basal operations. After analysis activities, the essential operations are named individually, and they constituent atomic operations. Descriptions of the atomic operations are listed in Table 1.
3.4. Inference Rules
All conventional reasoners use algorithms to infer implicitly stated knowledge and support basic reasoning tasks, e.g., determining if a given individual is an instance of a given concept. Systems usually support the following standard reasoning tasks:
Table 1 Atomic Operations
Atomic Operation | Description |
---|---|
Relation Search | Search following roles that work as relations among concepts in domain. |
Associate | Associate two concepts with a role works as relationship. |
Analogies | Look for similar concepts or roles. |
Perspective Change | Look for different concepts, different domains, or both. |
Force Fit | Force unrelated roles fit into concepts. |
However, this research requires more creative and complex queries. The required creative ideas cannot simply be searched from existing knowledge due to its creativity requirements: H-creative ideas with new, valuable, and surprising features. Therefore, new inference rules are necessary to support required semantic reasoning for idea elements generation.
As explained earlier, inference rules are fundamental for idea elements generation. Hence, we design a set of inference rules by adopting one or more atomic operations presented in Section 3.2. Because an operation does not necessarily support only one rule, the atomic operations are not categorised. Instead, the applications of the atomic operations are reflected in the designed rules for reasoning activities. Furthermore, the three types of creativity proposed by Boden [35], including exploration, transformation, and combination, are adopted as creativity principles, which are considered as the theoretical foundations of using the atomic operations to support the design of reasoning rules. Meanwhile, description logic is the fundamental logic of the inference rules. In short, specific inference rules are designed by adopting and combining the atomic operations following the three creativity principles, and they are organised based on description logic. Accordingly, the rules are expressed in the
Rule name: Analogism force-fitting
Where C, Dj, and E are different concepts and Ri is a specific role.
Rule (1) means the following:
Rule name: Relation force-fitting
Where C, Dj, and E are different concepts and Ri and Lm are specific roles.
Rule (2) means the following:
Rule name: Analogism new perspective
Where
Rule (3) means the following:
Rule name: Analogism relation association
Where
Rule (4) means the following:
Rule name: Relation transit association
Where
Rule (5) means: while
Five designed inference rules in formal presentations were presented above. They are crucial in semantic reasoning phase for idea creation. The rules are designed for general reasoning purposes; therefore, minor justification is allowed on the rules to develop domain related inference rules to support particular requirements on different applications.
The proposed method, especially the group of designed inference rules, is illustrated on a major recommendation test case. A prototype of creative major recommender (CMR) is designed and implemented to apply the proposed knowledge-based semantic reasoning method. In this application, recommended courses are considered as created ideas. This software system aims to assist students in exploring various options on choosing university major. Students' personal information are inputs, and the raw data for knowledge are information about universities and courses. The former is obtained from users' manual selection; the latter are crawled from UK universities' websites and manually organised and managed as an OWL file.
The architecture of the prototype software system is shown in Figure 4. It consists of a preprocessing and mapping module, an inference module, and a GUI and request module. Among them, the preprocessing and mapping module is performed offline in the server-end. The inference module interacts with the client-end GUI and request module. After receiving a user request, the major inference is executed on the domain knowledge base (KB), and the designed inference rules are implemented in this module; if there is no proper knowledge inferred, the keyword inference is executed on the common-sense knowledge base WordNet and the domain KB to infer a domain concept from a user keyword, while the keyword does not exist in the domain KB and the inferred concept is contained in the domain KB. The entire prototype is mainly implemented with the help of HTML, Java, SPARQL, and OWL.
5.1. Experimental Setup
According to the architecture design of the prototype software system, a major recommender prototype is implemented. For better user interaction with the system and the display of results, a client-end GUI is designed and implemented, and the interfaces are shown in Figure 5 and Figure 6. Particularly, in Figure 5, a set of inputs is selected for the experiment.
The outcomes of a set of input data are two conventional majors and two inspirational majors, as shown in Figure 6. The former is the suggestion that can be seen on other major recommenders, while the latter is the most likely unexpected result, and these inspirational recommendations distinguish this system from the others.
(a) (b)
Figure 6. (a) Conventional majors recommended by the CMR prototype; (b) Inspirational majors recommended by the CMR prototype
The conventional options are obtained by searching through the knowledge base to find the best matches between the ontology concepts and input keywords. The inspirational options are generated via the proposed semantic reasoning method, which enhances the capability of expanding other knowledge related to major choosing and considers creative elements as kernel factors. The results are evaluated in Section 5.2.
5.2. Evaluation Metrics
There is a set of evaluation metrics to measure creativity degrees including novelty, usefulness, and surprising [32]. The scores of the proposed three creativity elements, Novelty, Usefulness, and Surprising, are calculated for the recommended major options provided in the experiment. Furthermore, their overall creativity values are determined based on scores of the three creativity elements. The evaluation results are shown in Figure 7. Moreover, this section omits the specific calculation steps due to page limitation. To simplify this calculation, the differences between the weights of elements and sub-dimensions are all ignored.
(a) (b)
Figure 7. (a) Evaluated degrees of novelty, usefulness, and surprising; (b) Evaluated degrees of creativity
According to the evaluation results, it it shown that the conventional majors (option 1 and option 2) have strong usefulness as they are closely related to the inputs, and their degrees of novelty and surprising are relatively low. On the contrary, the inspirational majors (option 3 and option 4) gained high scores in novelty and surprising, especially surprising. Although the inspirational suggestions not directly linked to the inputs, the implicit relations between the course knowledge and the inputs are revealed by semantic reasoning. Accordingly, the inspirational major options obtained higher creativity degrees. In summary, the evaluation illustrates that the proposed knowledge-based semantic reasoning method works effectively at enhancing an idea's creativity.
To extend the creativity level in general idea creation, a knowledge-based semantic reasoning method is proposed in this paper. In particular, an idea creation framework is presented with three phases to provide the capability of being creative in idea generation. Most importantly, a set of inference rules are presented to support the semantic reasoning of idea creation. Atomic operations are defined based on creativity techniques, which is the fundamental actions of the inference rules and brings creativity features into the idea elements. On the basis of description logic and atomic operations, five inference rules are designed to support creativity in the proposed knowledge-based semantic reasoning. A prototype of creative major recommender illustrates the feasibility of the proposed semantic reasoning method, in which the created ideas are two conventional majors and two inspirational majors. The experiment results are evaluated by calculating the creativity degrees and comparing the scores of conventional options and inspirational options. The evaluation shows that the proposed semantic reasoning method can enhance an idea's creativity effectively.
The authors have declared that no competing interests exist.
[1.] |
|
[2.] |
|
[3.] |
|
[4.] |
“Managing Ideas for Profit: the Creative Gap (Marketing for Professionals),” McGraw-Hill, ,
|
[5.] |
|
[6.] |
|
[7.] |
“The Act of Creation,” Hutchinson, ,
|
[8.] |
|
[9.] |
“Applied Imagination: Principles and Procedures of Creative Problem-Solving,” 3rd Edition, Charles Scribner's Sons, ,
|
[10.] |
|
[11.] |
|
[12.] |
Jogalekar and A. Mangla, “Idea Generation Algorithm based Systems,” in .
|
[13.] |
|
[14.] |
|
[15.] |
|
[16.] |
|
[17.] |
|
[18.] |
|
[19.] |
|
[20.] |
|
[21.] |
|
[22.] |
|
[23.] |
|
[24.] |
|
[25.] |
|
[26.] |
|
[27.] |
|
[28.] |
|
[29.] |
“A Structural Paradigm for Representing Knowledge (BBN Report No. 3605),
|
[30.] |
|
[31.] |
|
[32.] |
|
[33.] |
|
[34.] |
|
[35.] |
“The Creative Mind: Myths and Mechanisms,” 2nd Edition, Routledge, ,
|
/
〈 | 〉 |