IngeniousTRIZ: An automatic ontology-based system for solving inventive problems

Knowledge-Based Systems xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Knowledge-Based Systems

journal homepage: www.elsevier .com/ locate /knosys

IngeniousTRIZ: An automatic ontology-based systemfor solving inventive problems

http://dx.doi.org/10.1016/j.knosys.2014.11.0150950-7051/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author at: School of Information Science and Engineering,Shandong Normal University, Jinan City, China. Tel.: +86 15165125108.

E-mail addresses: [email protected] (W. Yan), [email protected] (H. Liu),[email protected] (C. Zanni-Merk), [email protected] (D. Cavallucci).

1 The contradiction in TRIZ are divided into: technical contradiction andcontradiction.

Please cite this article in press as: W. Yan et al., IngeniousTRIZ: An automatic ontology-based system for solving inventive problems, Knowl. Base(2014), http://dx.doi.org/10.1016/j.knosys.2014.11.015

W. Yan a,b,⇑, H. Liu a,b, C. Zanni-Merk c, D. Cavallucci d

a School of Information Science and Engineering, Shandong Normal University, Jinan City, Chinab Shandong Provincial Key Laboratory for Novel Distributed Computer Software Technology, Jinan City, Chinac ICUBE/BFO Team (UMR CNRS 7357) – Pole API BP 10413, Illkirch 67412, Franced LGECO/INSA Strasbourg, 24 Boulevard de la Victoire, Strasbourg 67084, France

a r t i c l e i n f o

Article history:Received 4 June 2014Received in revised form 1 October 2014Accepted 14 November 2014Available online xxxx

Keywords:The theory of inventive problem solving(TRIZ)Knowledge sourcePhysical–chemical–geometrical effectsSemantic similarityOntologyOntology reasoning

a b s t r a c t

With the development of the Theory of Inventive Problem Solving (TRIZ), different models and knowl-edge sources were established in order to solve different types of inventive problems. These knowledgesources with different levels of abstraction are all built independently of the specific application field, andrequire extensive knowledge about different engineering domains. In order to facilitate the use of theseTRIZ models and knowledge sources, an intelligent knowledge management system – IngeniousTRIZ wasdeveloped in this research. On the one hand, according to the TRIZ knowledge sources ontologies, thissystem offers to the users the relevant knowledge sources of the model they are building. On the otherhand, the system has the ability to fill ‘‘automatically’’ the models of the other knowledge sources. Firstly,the TRIZ user chooses a TRIZ knowledge source to work for an abstract solution. Then, the items of otherknowledge sources, which are similar with the selected items of the first knowledge source, are obtainedbased on semantic similarity calculated in advance. With the help of these similar items and the heuristicphysical effects, other concept solutions are generated through ontology inference. In order to show thiswhole process, the resolution of the case of the ‘‘diving fin’’ in IngeniousTRIZ system is elaborated indetail.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

The Theory of Inventive Problem Solving (TRIZ) was developedby G. Altshuller in Russia from 1946 to 1985. The goal of this meth-odology was, initially, to improve and facilitate the resolution oftechnological problems [2,3]. The basic idea in TRIZ is that (techni-cal) systems evolve in similar ways, and by reducing any situationand problem to a physical level, standard solutions and problemsolving techniques, borrowed from many different fields, can beapplied. In order to be also suited to the complex inventive prob-lems, several extensions were proposed, such as OTSM-TRIZ orInventive Design Methodology (IDM) [6,23].

Fig. 1 describes the way classical TRIZ solves problems, and theinterrelations among its different parts. Three different phases areclearly identified:

� The ‘‘formulation’’ phase, where the expert uses different toolsto express the problem in the form of a contradiction1 networkor another model.� The ‘‘abstract solution finding’’ phase, where access to different

knowledge bases is made to get one or more solution models.Generally, in this step, TRIZ users are required to have wideexperience on the TRIZ knowledge sources (Section 3). Theyneed to be capable of choosing the accurate abstract solutionaccording to the current abstract problem.� The ‘‘interpretation’’ phase, where these solution models are

instantiated with the help of the scientific-engineering effectsknowledge base, to get one or more solutions to be imple-mented in the real world.

According to classical TRIZ, the resolution of inventive problemsconsists in the construction of models and the use of the corre-sponding knowledge sources. As shown in Fig. 1, different modelsand knowledge sources were established in order to solve different

physical

d Syst.

http://dx.doi.org/10.1016/j.knosys.2014.11.015

mailto:[email protected]



mailto:denis.cavallucci@insa-stras bourg.fr

mailto:denis.cavallucci@insa-stras bourg.fr


http://www.sciencedirect.com/science/journal/09507051

http://www.elsevier.com/locate/knosys


Abstract problem

Abstract solution

Physical contradiction

Technical contradiction

Substance-Field model

FORMULATION

Specific inventive technical problem

Specific inventive technical conceptual

solution

ABSTRACT FIELD

Laws of technological

innovation

Physical, chemical,

geometrical effects

40 inventive principles11 separation principles76 inventive standards

(From the roughest model to the most well informed model)

Abstract

Concrete

INSTANTIATION

Fig. 1. The inventive problem solving approach.

From the roughest model to the most well informed model

From the most abstract knowledge to the most concrete model

Fuzzy Model

Technical Contradiction Model

Physical Contradiction Model

Substance-Field Model

Inventive Principles

Separation Methods

Inventive Standards

Physical Effects

Fig. 2. The main successive models in TRIZ.

2 W. Yan et al. / Knowledge-Based Systems xxx (2014) xxx–xxx

types of inventive problems, such as 40 inventive principles(Section 3.1) for eliminating the technical contradictions2 and 11separation methods (Section 3.3) for eliminating the physical con-tradictions.3 However, there exist several problems in the use ofthese models and knowledge sources:

� There are different models at different levels of abstraction. Asshown in Fig. 2, the initial problem being often fuzzy and notclearly expressed, the methodology pushes the designer to thebuilding of a model in terms of a set of predefined TRIZ con-cepts. This model is called a ‘‘systemic model’’. Then, a secondmodel emphasizes the main contradiction that is at the baseof the problem (‘‘technical contradiction’’ or ‘‘physicalcontradiction’’).� Each model is built independently of the specific application

field and has its set of associated knowledge sources. The reso-lution with different models is done starting from the differentassociated knowledge sources.� There exists a gap between the high-level abstraction of TRIZ

and reality, and TRIZ does not give directives on the way thesemodels have to be used.

In order to facilitate this innovative process, an intelligentknowledge management system – IngeniousTRIZ was developedin this research. On the one hand, according to the TRIZ knowledgesources ontologies, this system offers to the users the relevantknowledge sources of the model they are building. On the otherhand, the system has the ability to fill ‘‘automatically’’ the modelsof the other knowledge sources.

The remainder of the paper is organized as follows. Section 2presents the state of the art about existing systems and softwareto cope with similar problems in TRIZ, which proves the necessityof this research. In Section 3, the main TRIZ knowledge sources are

2 A technical contradiction arises when it is required to improve some feature ofthe existing prototype but all solutions known within the domain do not produce therequired result or their use would cause a negative effect.

3 A physical contradiction indicates that a part of a design prototype should havetwo mutually exclusive values of the same physical parameter.

Please cite this article in press as: W. Yan et al., IngeniousTRIZ: An automatic(2014), http://dx.doi.org/10.1016/j.knosys.2014.11.015

introduced. A general introduction about physical–chemical–geometrical effects is given in Section 4. Section 5 discusses theproposed methods in detail. Section 6 gives the framework ofIngeniousTRIZ system and elaborates the whole process of usingit to solve the specific case of the ‘‘diving fin’’. A short discussionis included in Section 7 and finally, the conclusions and directionsfor future research are presented in Section 8.

2. State of the art

In recent years, several different software and databases, weredeveloped to automate and facilitate the process of using TRIZ.

The software STEPS, developed by LGECO/INSA de Strasbourg, isable to effectively accompany the user from the formulation phaseof the problem to its resolution, taking into account the specificcontext of its application. STEPS exists in two versions: A commer-cial one, managed by a start-up company, Time To Innovate4; and aresearch version, property of the consortium and INSA de Strasbourg,

4 http://www.time-to-innovate.com.

ontology-based system for solving inventive problems, Knowl. Based Syst.

http://www.time-to-innovate.com


W. Yan et al. / Knowledge-Based Systems xxx (2014) xxx–xxx 3

which implements the ontology of the TRIZ/IDM5 concepts whichhave been created in [22,23]. Both versions of STEPS can providevaluable assistance to the user through the intelligent analysis basedon the statistic data. However, using STEPS to solve a specific prob-lem requires extensive knowledge of different engineering domainsand is not currently supported by it. Consequently, the resolutionmainly depends on the experience and knowledge of the user.

Invention Machine’s Goldfire,6 a commercial software productfrom Invention Machine, provides tools for the organization’s man-agement of technology and for the individual’s researching and engi-neering tasks. The scientific effects tool helps to stimulate creativeproblem solving by browsing and searching effects in the InventionMachine Scientific Effects Database. The physical effects are dividedinto 26 classes according to different functions, such as, ‘‘Substance:Absorb/Adsorb’’, and then for each class, there are several sub-clas-ses, for example, for the class ‘‘Substance: Absorb/Adsorb’’, thesub-classes are ‘‘absorb/adsorb gas’’, ‘‘absorb liquid substances’’,etc.Each sub-class also consists of several sub-subsidiary classes, forexample, ‘‘absorb/adsorb gas’’ is made up of ‘‘absorb/adsorb gas’’and ‘‘absorb/adsorb vapor’’. As a result, the user needs to locatethe specific case level-by-level to obtain the physical effects. Withthe good performance of the proposed semantic methods and thestandard representative way of problem and solution, the resultsare almost satisfied. However, the semantic search, only dependingon the matching among two words with the same role in the sen-tence, cannot provide more extensive semantic information, suchas, a physical effect related to ‘‘absorb vapor’’ is also an effect to‘‘absorb gas’’.

The CREAX database7 organizes a database of effects by func-tion, and uses a web-based application to support the search ofeffects. It consists of two databases, that is, the Function Databaseand the Attribute Database. In the Function Database, the userneeds to choose a Function(Pointers) and an object, which are pre-defined in advance, such as, ‘‘Absorb Gas’’, and then all the relatedphysical effects are obtained. In the Attribute Database, the useralso needs to select an Attribute and a kind of behavior, such as,‘‘Changing Colour’’, to search for the appropriate physical effects.However, this system responds to user-entered query by processingthe query to a combination of two prestored key words, which isquite limited in the specific applications. Furthermore, the classifi-cation of objects – ‘‘Solid’’, ‘‘Liquid’’, ‘‘Gas’’ and ‘‘Field’’ is at highlevel of abstraction, and too many physical effects are eligible fora specific application.

As stated above, the process of using existing software still com-pletely or partly depends on users, requiring a high expertise inTRIZ usage to appropriately manipulate these concepts. Mean-while, in order to represent the concepts and the conceptual archi-tecture, ontology has been gradually incorporated to supportknowledge management systems in different domains. For exam-ple, Amailef and Lu [4] proposed an ontology-supported case-based reasoning(OS-CBR) method to support emergency decisionmakers to effectively respond to emergencies, Gil and Martin-Bau-tista [11] proposed a novel model of an Ontology-Learning Knowl-edge Support System (OLeKSS), and Rodríguez-García et al. [16]presented an ontology-based platform which can assist to matchuser’s needs in the process of discovering the cloud services. As aresult, this research intends to build an ontology-based intelligentknowledge management system to facilitate the use of TRIZ.

5 IDM (Inventive Design Methodology) [22,23] is an extension to TRIZ, that isintended to process problems with hundreds of parameters, contradictions andproblems not necessarily linked to inventive product design (for example, in softwareengineering or in the organization field).

6 http://inventionmachine.com/products-and-services/innovation-software/.7 http://function.creax.com/.


3. The TRIZ knowledge sources

The knowledge sources8 for solving inventive design problemsincluding 40 inventive principles, 76 inventive standards and 11 sep-aration methods, are used to eliminate technical contradictions, pro-vide common problem-solving methods, such as Su-Field analysis,and eliminate physical contradictions respectively.

3.1. Inventive principles

They are heuristic principles based on the accumulated andgeneralized previous experience of inventors. Due to a high degreeof generalization, they are available in a form that is independentof any particular engineering domain. TRIZ formulates 39 GenericEngineering Parameters (GEPs), like ‘‘the weight of a movableobject’’ or ‘‘speed’’.

The inventive principles can be used in a systematic way byaccessing the principles through indices in the contradictionmatrix. Along the vertical axis of this matrix the GEPs which haveto be improved are specified while along the horizontal axis theGEPs which deteriorate as a result of the improvement are speci-fied. These parameters can be looked up along the vertical and hor-izontal axes and the matrix suggests up to 4 principles that can beused to solve the contradiction without causing negative effect. Anexample is:

� Inventive Principle 35: Change of physical and chemicalparameters.(a) Change the object’s aggregate state.(b) Change concentration or consistency of the object.(c) Change the degree of flexibility of the object.(d) Change the temperature of the object or environment.

3.2. Inventive standards

They are drawn from the fact that most inventions refer to con-ceptual modification of physical systems. If problems from differ-ent domains result in identical physical models, this means thatthe problems are similar. Therefore, they can be solved by applyingthe same method. Standards are built in the form of recommenda-tions, and generally, formulated as rules like If hCondition1i andhCondition2i then hRecommendationi. Both conditions permit rec-ognizing the typology of the problem associated to the standard.This way, for a built problem model, there exist a certain numberof recommendations allowing the construction of the correspond-ing solution model. In TRIZ, 76 inventive standards are available. Aproblem with these inventive standards is that they are formulatedabstractly, so their practical use is quite difficult. An example is:

� Inventive Standard 5.3.1: Changing of phase state.(a) Efficiency of the use of substance without introducing other

substances is improved by changing its phase.

3.3. Separation methods

An advanced form of contradictions is the physical contradic-tions [2]. To model an inventive problem as a physical contradic-tion, a physical object of a prototypical design that must havetwo conflicting properties has to be identified. To solve problemscontaining physical contradictions, separation methods for physi-cal contradiction elimination are used. Among them, there are:

8 The whole texts of TRIZ knowledge sources can be found in http://www.time-to-innovate.com/.


http://inventionmachine.com/products-and-services/innovation-software/

http://function.creax.com/

http://www.time-to-innovate.com/

http://www.time-to-innovate.com/



� Separation of conflicting properties in time.� Separation of conflicting properties in space.� Separation of conflicting properties by system transition.� Separation of conflicting properties by phase transition.� Separation of conflicting properties by physical–chemical

transition.

4. Physical–chemical–geometrical effects

While inventive principles, separation methods and inventivestandards do not produce recommendations in terms of whatphysical substances or fields should be used, the collection of phys-ical–chemical–geometrical effects (also called physical effects forshort) provides the mapping between technical functions andknown natural laws.

Example: Instead of a mechanical design including many partsfor the precise displacement of an object for a short distance, it ispossible to apply the effect of thermal expansion to control it.

As shown in Fig. 3, the technical functions, also called pointersto physical effects, are defined to represent the way to use theseeffects. For example, if the shape of the system needs to be chan-ged, that is, the technical function of this system can be definedas ‘‘change shape’’, and then through this pointer, several usefulphysical effects can be obtained, such as, ‘‘Evaporation’’ and‘‘Crystalization’’.

5. Proposed methods

In this section, three core approaches, that is, semantic similar-ity calculation, ontology-based knowledge modeling and ontologyinference, are described in full details.

5.1. Semantic similarity calculation

Semantic similarity is a measure of the likeness of meaning orsemantic content assigned to a set of documents or to terms withinterm lists. In recent years, many research has been done on themeasures of semantic similarity, for example, in order to discoversemantic correspondences between concepts defined in DTDsand XSDs, many XML-specific Schema Matching approaches, calledXML Matchers, are designed[1], and in order to provide vastamounts of common-sense and domain-specific knowledge forcomputing semantic similarity, Gabrilovich and Markovitch [10]proposed to represent the meaning of texts in a high-dimensionalspace of concepts derived from Wikipedia. Taking into account thespecific applications in TRIZ, we are interested in the measures thatuse a thesaurus, such as WordNet [8]. The methods developed tomeasure semantic similarity based on thesaurus can be catego-

Fig. 3. The organization of the pointers to the physical–chemical–geometricaleffects.


rized in three groups: Path-based Similarity, Information Content(IC) Similarity and Extended Gloss Overlaps [15]. According tothe characteristics of the items to be matched, such as, highappearance of words with similar meaning, Lin’s method [14] ischosen to calculate the sense similarity, based on which the wordsimilarity and short-text similarity are calculated.

5.1.1. Basic method for matching SPs and GEPsAs presented in Section 3.1, in order to solve the specific contra-

diction, two Specific Parameters (SPs) firstly need to be matchedwith appropriate GEPs respectively. However, in real-world prob-lems, most of the times, contradictions are established in termsof parameters that are inherent to the artefact that is being devel-oped, and there is a semantic gap to fill between those parametersand the generalized ones. An abstraction effort needs to be pro-vided to choose the best GEP, and in this way, be able to use thecontradiction matrix.

In order to facilitate this process, a method to calculate thesemantic similarity between short texts is explored to fill the gapbetween SPs and GEPs:

� Preprocess: It mainly includes two steps: Subjective WordsDeletion and Semantic Extension. In the first step, the uselesssubjective words (often noun, and are used to represent specificobjects) in SPs are eliminated. And then, the useful wordsobtained in the first step are extended with their similar wordsbased on WordNet.� The matching process: For two sentences to be matched, firstly

they are divided into several words by using sentence segmen-tation (e.g., use Decision Tree to determine a word is end of sen-tence or not), word tokenization (e.g., substance appearance –disappearance ! hsubstance, appearance, disappearancei), andword normalization and stemming (e.g., automates, automatic,automation ! automat). For each word obtained, WordNet isused to look for its corresponding senses, including nouns,verbs, adjectives and adverbs. Then, in order to calculate thesemantic similarity between two senses, Lin’s measure is used,and the maximum sense similarity of the two words is definedas their word similarity. Finally, short text similarity is calcu-lated based on word similarity as following:

sðA1;A2Þ ¼Pn

i¼1 Maxmj¼1ðsðA1i;A2jÞÞ

nð1Þ

where A1 is the first short text to be matched, including wordssequence A11; A12 � � �A1n and A2 is the second short text to bematched, including A21; A22 � � �A2m. sðA1i;A2jÞ represents wordsimilarity of A1i and A2j; 1 6 i 6 n; 1 6 j 6 m.

5.1.2. Improved method for defining the missing links among the TRIZknowledge sources

As shown in Section 3, the knowledge sources for resolutionsare situated at different levels of abstraction and at different levelsof ‘‘closeness to reality’’ [5]. For example, for inventive principles,even if they seem to refer to concrete reality (for instance, ‘‘inertatmosphere’’) are conceptually more abstract than inventive stan-dards, which refer to concrete substances or fields.

More than that, even in the same knowledge source the descrip-tion may be different or incompatible, as stated in [6]. For example,even though both Inventive Standard 1.1.5 – ‘‘Transition to SFM byusing external environment with additives’’ and Inventive Standard1.1.3 – ‘‘Transition to external complex SFM’’ indicate bring in theexternal substances or fields to improve the existing model, theirdescription are different.

In this research, the method of measuring short-text semanticsimilarity are used to compare, analyze, and match the items in



Inventive Standard 1.1.5: Transition to SFM by using external environment with additives

Inventive Standard 1.1.3: Transition to external complex SFM

Inventive Standard 5.3.1: Changing of phase state

Inventive Principle 35: Change of physical and chemical parametersInventive Principle 39: Inert atmosphere

Sim: 0.68 Sim: 0.84

Sim: 0.33

Sim: 0.51

Sim: 0.73Sim: 0.75Sim: 0.78 Sim: 0.77

Sim: The value of similarity (The larger the calculated similarity, the more similar the two short texts).

Sim: 0.15 Sim: 0.05

Fig. 4. An example of correspondences of items of different knowledge sources.


the different knowledge sources ontologies. Fig. 4 shows an exam-ple of correspondences of items in the knowledge sources. Thehigher the value of similarity, the more similar the two items.For example, Inventive Standard 5.3.1 – ‘‘Changing of phase state’’is more similar to Inventive Standard 1.1.5 – ‘‘Transition to SFMby using external environment with additives’’ (Sim: 0.75) thanInventive Principle 39 – ‘‘Inert atmosphere’’ (Sim: 0.15).

The process of using semantic similarity to define the missinglinks among the TRIZ knowledge sources is made up of:

� Preprocess: Before the semantic matching, data preprocess iscarried out to reduce redundancy and minimize imprecision.Different kinds of knowledge sources need different preprocess-es, for example, the additional information, such as the similarwords, need to be added to inventive principles to facilitatetheir matches with inventive standards, while inventive stan-dards need to be classified into several kinds according to theircharacteristics.9

� The matching process: The basic matching approach asserts thatall the words in the short-text play an equal role in calculatingthe semantic similarity. According to the analysis of the docu-ments of TRIZ knowledge sources, words act differently on theshort text similarity, for example, the word ‘‘transition’’ canbe used to separate several short texts from others, while thewords ‘‘a’’ or ‘‘an’’ are useless. As a result, an improved methodis proposed with the word weight calculated by tf ⁄ idf, whichwere proposed by Salton and Lesk [17] and Sparck-Jones [18]respectively. Given two short texts A1 and A2, the most similarwords in A2 for each word in A1 are selected. However, it isnot enough for us to estimate the similarity between A1 andA2 without considering the inverse situation, that is, obtainthe most similar words in A1 for each word in A2. Taking thisinto account, the semantic matching with word weight isproposed:

9

and

Ple(20

sðA1;A2Þ ¼Pn

i¼1ww1i �Maxmj¼1ðsðA1i;A2jÞÞ

n

þPm

i¼1ww2i �Maxnj¼1ðsðA2i;A1jÞÞ

mð2Þ

where sðA1i;A2jÞ represents word similarity of A1i andA2j; 1 6 i 6 n; 1 6 j 6 m, and sðA2i;A1jÞ represents word similar-ity of A2i and A1j; 1 6 i 6 m; 1 6 j 6 n, calculated as stated inbasic matching. ww1i: the word weight of the ith word in A1,and ww2i: the word weight of the ith word in A2. Both of themcould be calculated by using the methods tf � idf .

71 standards are identified as the implementation of existing inventive principles,the others are identified as new inventive principles and patterns of evolution.

ase cite this article in press as: W. Yan et al., IngeniousTRIZ: An automatic14), http://dx.doi.org/10.1016/j.knosys.2014.11.015

5.2. Ontology-based knowledge modeling

An ontology is composed of concepts and relationships that areused to express knowledge about the modeling field [12]. There-fore, the constitution of a framework for problem-solving basedon different knowledge sources and physical effects as the aim ofontology will be considered.

In recent years, a lot of ontology languages have been developedand can be grouped into two major categories, that is, the lan-guages derived from knowledge representation paradigms in AIcommunity, such as KIF (Knowledge Interchange Format),10 andthe web-based ontology languages, such as RDF(S) (ResourceDescription Framework (Schema))11 and OWL (Web Ontology Lan-guage) [7].

In this research, three ontologies are built until now, that is, theinventive principles ontology, the inventive standards ontologyand the physical effects ontology.

Inventive principles ontology: The inventive standards ontologyis built to represent the terms and their relationships used in theprocess of using contradiction matrix and inventive principles.The framework of the inventive principles ontology is shown inFig. 5. The interpretations from AppliedFeature to PrimaryFeature,and from PrimarySubIP to AppliedSubIP need to be implementedmanually; and this fact requires a large amount of TRIZ experienceand knowledge covering a wide spectrum of domains. As shown inSection 5.1.1, this research uses semantic similarity to automate, asmuch as possible, this process in order to provide assistance toTRIZ users [19,20].

Inventive standards ontology: As shown in Fig. 6, the inventivestandards ontology is developed to represent the knowledge usedin Sub-Field analysis.

The physical effects ontology: Fig. 7 shows the framework of thephysical effects ontology. In this ontology, the concepts and theirrelationships about the use of physical effects are constructed.The class Physical_Effect is built with the property keyword, whichmakes it possible to obtain the heuristic physical effects throughthe search of keyword [21].

5.3. Ontology inference

5.3.1. Ontology reasoning rulesSemantic Web Rule Language (SWRL), which combines sub-

languages of the OWL (OWL-DL and OWL-Lite) with the RuleMarkup Language,12 is used to describe the reasoning rules in thisautomatic process of solving inventive problems.

10 http://logic.stanford.edu/kif/dpans.html.11 http://www.w3.org/TR/1998/WD-rdf-schema-19980409/.12 http://www.ruleml.org.


http://logic.stanford.edu/kif/dpans.html

http://www.w3.org/TR/1998/WD-rdf-schema-19980409/

http://www.ruleml.org


-version: StringContradiction Matrix

139*39

hasPositivePrimaryFeature

hasNegativePrimaryFeature-description: String

PrimaryFeature

-description: String-hasPositiveAppliedFeature: AppliedFeature-hasNegativeAppliedFeature: AppliedFeature-hasSolution: AppliedSubIP

Problem

hasPositiveAppliedFeature

hasNegativeAppliedFeature

-IPNum: IntInventivePrinciple

10...n

-IPNum: Int-SubIPNum: Int-hasPrimaryDescriptionOfSubIP: PrimarySubIP-hasAppliedDescriptionOfSubIP: AppliedSubIP-hasSubject: Subject-hasPredicate: Predicate-hasObject: Object

SubInventivePrinciple

11...n

-description: String-hasSubject: Subject-hasPredicate: Predicate-hasObject: Object

PrimarySubIP

-description: String-hasSubject: Subject-hasPredicate: Predicate-hasObject: Object

AppliedSubIP

hasPrimaryDescriptionOfSubIP

hasAppliedDescriptionOfSubIP hasSolution

-hasPositivePrimaryFeature: PrimaryFeature-hasNegativePrimaryFeature: PrimaryFeature

Item-description: String

AppliedFeature

-description: StringFeature

Fig. 5. The inventive principles ontology.

-description: String-correspondsTo_GPM: Generic_Problem_Model-correspondsTo_GSM: Generic_Solution_Model-chooses_IS: InventiveStandard-chooses_PE_Type: int-chooses_PE: Physical_Effect

Problem

-description: String-hasSubstance1_GPM: Substance-hasSubstance2_GPM: Substance-hasField_GPM: Field-has_S1_Num_GPM: int-has_S2_Num_GPM: int-has_F_Num_GPM: int

Generic_Problem_Model

-description: String-hasSubstance1_GSM: Substance-hasSubstance2_GSM: Substance-hasField_GSM: Field-has_S1_Num_GSM: int-has_S2_Num_GSM: int-has_F_Num_GSM: int-changesOnOrFor_Element: Element

Generic_Solution_Model

-description: String-includes_Sub: Substance

Substance-description: String-includes_Field: Field

Field

-description: String-has_keyid_of_type: int

InventiveStandardcorrespondsTo_GPM correspondsTo_GSM

hasSubstance1_GSM / hasSubstance2_GSM

hasField_GSMhasSubstance1_GPM / hasSubstance2_GPM

hasField_GPM

Choose_IS

chagesOnOrFor_Element

-description: String-includes: Element

Element

Fig. 6. The inventive standards ontology.


A SWRL rule axiom is made up of an antecedent and a conse-quent, both consisting of one or more atoms. In a rule with multi-ple atoms in the body, the body is treated as a conjunction of its


atoms. In a rule with multiple atoms in the head, the head also istreated as a conjunction. But such a rule can be easily transformedinto multiple rules each with a single-atom head [13].



Fig. 7. The physical effects ontology.


The rules in this application can be divided into two kinds:

� 40 rules for searching heuristic abstract solutions: These rulesare defined to automate the process of using inventive stan-dards to obtain the heuristic abstract solutions. According tothe analysis of 76 standards, the number of substances andfields can be used to identify the modification of inventive stan-dards. So the properties has_S1_Num_GPM, has_S2_Num_GPM,has_F_Num_GPM are used to indicate the number of substancesand field before the modification while has_S1_Num_GSM,has_S2_Num_GSM and has_F_Num_GSM are used to indicatethe number of substances and field after the modification. Theregular patterns about the number of substance and field aredescribed as following:

– For Substance 1:

Please(2014)

⁄ has_S1_Num_GPM – 0:

cite th, http:/

� has_S1_Num_GSM = 1: The additive Substance 1 isused instead of the Substance 1 before themodification.

� has_S1_Num_GSM – 1: The additive Substance 1 isadded to the Substance 1 before the modification.

⁄ has_S1_Num_GPM = 0:
� The additive Substance 1 is added to complete the
incomplete problem model.

– For Substance 2:
⁄ has_S2_Num_GPM – 0:
Table 1SWRL rule for searching heuristic abstract solutions.

Name SWRL rule

Rule 1 Problemð?xÞ ^ InventiveStandardð?yÞ ^ choose ISð?x; ?yÞ^

� has_S2_Num_GSM = 1: The additive Substance 2 isused instead of the Substance 2 before themodification.

� has_S2_Num_GSM – 1: The additive Substance 2 isadded to the Substance 2 before the modification.

has keyid of typeð?y;1Þ ^ Generic Problem Modelð?zÞ^correspondsTo GPMð?x; ?zÞ ^ Generic Solution Modelð?aÞ^
⁄ has_S2_Num_GPM = 0: correspondsTo GSMð?x; ?aÞ ^ Substanceð?cÞ ^ Substanceð?dÞ^hasSubstance1 GPMð?z; ?cÞ ^ hasSubstance2 GPMð?z; ?dÞ^has S1 Num GPMð?z;1Þ ^ has S2 Num GPMð?z;1Þ^
� The additive Substance 2 is added to complete theincomplete problem model.

has F Num GPMð?z;0Þ ! hasField GSMð?a; added F z2Þ^hasSubstance1 GSMð?a; ?cÞ ^ hasSubstance2 GSMð?a; ?dÞ^
has S1 Num GSMð?a;1Þ ^ has S2 Num GSMð?a;1Þ^ - For Field: has F Num GSMð?a;1Þ ⁄ has_F_Num_GPM – 0:
is article in press as: W. Yan et al., IngeniousTRIZ: An automatic ontology-based/dx.doi.org/10.1016/j.knosys.2014.11.015

� has_F_Num_GSM = 1: The additive Field is usedinstead of the Field before the modification.

� has_F_Num_GSM – 1: The additive Field is addedto the Field before the modification.

⁄ has_F_Num_GPM = 0:
� The additive Field is added to complete the incom-
plete problem model.

� 50 rules for searching heuristic physical effects: These rules areexplored to automate the process of searching heuristic physi-cal effects, and are divided into two classes: 42 IS (InventiveStandard) rules and 8 PE (Physical Effect) rules. The inferencewith the IS rules yields to several abstract types of physicaleffects, and the PE rules are used to find the concrete physicaleffects to instantiate the solution model.

Table 1 shows an example of rules for searching heuristicabstract solutions. Through the inference with this rule, the miss-ing Field will be added to the problem model.

5.3.2. Ontology reasoning engineJess (Java Expert System Shell) is a rule engine for the Java plat-

form, developed by Ernest Friedman-Hill of Sandia National Labssince 1995 [9]. Jess supports the development of rule-based expert

system for solving inventive problems, Knowl. Based Syst.


System: Ingenious TRIZ

OWL-API

Instances of OWL OntoBuild

OWL OntoAsserted and Inferred

parts

ReasonOWL Onto and

SWRL rules

UseProtege-OWL API

Return

SWRL API

Is-a

SWRLRuleEngineBridge

SWRL Bridge

Is-aSWRLJessBridge

SWRL Rule Engine API

Include

Include

Fig. 8. The interaction between IngeniousTRIZ and ontologies.

Step1 Contradic�on Descrip�on

Step1 Contradic�on Descrip�on

Step2 Contradic�on Formaliza�on

Step2 Contradic�on Formaliza�on

Step3 SP-> GEPStep3 SP-> GEPStep5 Seman�c

Inven�ve Standard

Step5 Seman�c Inven�ve Standard

Step6 Su-Field Analysis-1




Step9 Heuris�c Physical Effects-1



Heuris�c Concept Solu�on-3

Concept Solu�on-1

Heuris�c Concept Solu�on-2



Step4 Inven�ve Principle

Step4 Inven�ve Principle

SP: Specific ParameterGEP: Generic Engineering ParameterPhase 1

Phase 2Phase 3

Fig. 9. The framework of IngeniousTRIZ.


systems which can be tightly coupled to code written in the pow-erful, portable Java language.

The inference process of Jess includes three steps:

� Step 1: OWL + SWRL ! Jess will transfer the appropriate OWLknowledge and the rule knowledge in SWRL to Jess.� Step 2: After importing Jess files containing both Jess facts and

Jess rules, the inference process is implemented by executingJess.� Step 3: The Jess facts, including asserted facts and inferred facts,

are back to OWL knowledge.

6. IngeniousTRIZ system

The IngeniousTRIZ system was developed in a Java 1.7.02 plat-form, MySQL 5.1.22, WordNet 2.0, Protégé 3.4.3 and Jess 7.1p2 on aWindows environment. In order to manipulate the OWL ontologiesin the Java application, two inferfaces – Protege-OWL API13 and OWLAPI14 are used.

As shown in Fig. 8, OWL-API is used to build the instances andtheir properties, SWRL Rule Engine API15 to execute the ontology

13 http://protegewiki.stanford.edu/wiki/ProtegeOWL_API_Programmers_Guide.14 http://owlapi.sourceforge.net/index.html.15 SWRL Rule Engine API is packaged with Protege-OWL API before Protégé 3.5

(included).


reasoning, and Protege-OWL API to read the inferred results, whichare stored in new ontologies after the inference.

6.1. Framework

As shown in Fig. 9, the window-based IngeniousTRIZ consists ofthree continuous phases, and each part includes several steps asfollowing:

� Phase1: Step1!Step4 The resolution with contradiction matrix andinventive principles: The appropriate inventive principles can beobtained for the specific cases.� Phase2: Step4!Step8 Heuristic abstract solution from inventive

standards: The heuristic abstract solutions can be obtained inthis phase.� Phase3: Step8!Step11 The search of heuristic physical effects: The

heuristic physical effects can be obtained at the end of thisphase, based on which the concept solutions will be explored.

6.2. Case study: The case of the ‘‘diving fin’’

In order to illustrate the functions of IngeniousTRIZ system, thedetailed process of using it to solve the specific problem of the‘‘diving fin’’ is introduced in this section.


http://protegewiki.stanford.edu/wiki/ProtegeOWL_API_Programmers_Guide

http://owlapi.sourceforge.net/index.html


Fig. 10. The similar GEPs for SP-‘‘ease of use’’.

Fig. 11. The obtained inventive principles.


6.2.1. The problemEven if the use of diving fins becomes popular, there are still

some problems with them. The divers need to make a great effortto push water, which often makes them tired. Generally, the sur-face of the diving fin should be soft for offering minimal resistanceto water in order to minimize the effort of the diver and, at thesame time, it should also be hard in order to push water moreefficiently.

6.2.2. The resolution with contradiction matrix and inventiveprinciples

In this phase, the user works with contradiction matrix andinventive principles to obtain the abstract solution.

Firstly, the user needs to provide the parameters and values toformalize the technical contradiction16 for the case of the ‘‘divingfin’’. There is a contradiction between ‘‘Kicking efficiency’’ and ‘‘Easeof use’’, that is, if the surface of the diving fin is hard, it is better forthe kicking efficiency but worse for the ease of use, and vice versa.

In order to use contradiction matrix, SPs need to be matchedwith 39 GEPs. Based on the method proposed in Section 5.1.1, 5most similar GEPs are obtained through the semantic search foreach SP. For example, Fig. 10 shows the similar GEPs for SP-‘‘easeof use’’.

The two SPs in the case of the ‘‘diving fin’’ correspond to 5 sim-ilar GEPs respectively:

� SP1: kicking efficiency– GEP9: speed.– GEP19: use of energy by moving object.– GEP33: ease of operation.– GEP34: ease of repair.– GEP39: productivity.� SP2: ease of use

– GEP19: use of energy by moving object.– GEP20: use of energy by stationary object.– GEP32: ease of manufacture.– GEP33: ease of operation.– GEP34: ease of repair.

16 The process of defining the technical contradiction for a specific case isintroduced in the website: http://www.time-to-innovate.com/steps_matrix.


Taking into account the characteristics of this case, ‘‘SP1: kick-ing efficiency’’ is matched with ‘‘GEP19: use of energy by movingobject’’ while ‘‘SP2: ease of use’’ is matched with ‘‘GEP33: ease ofoperation’’.

Based on the two selected GEPs, a series of inventive principlesare obtained as shown in Fig. 11, and the detailed information is:

� Inventive Principle 32: Color changes(a) Change the color of an object or its external environment.(b) Change the transparency of an object or its external

environment.� Inventive Principle 28: Mechanics substitution

(a) Replace a mechanical means with a sensory (optical, acous-tic, taste or smell) means.

(b) Use electric, magnetic and electromagnetic fields to interactwith the object.

(c) Change from static to movable fields, from unstructuredfields to those having structure.

(d) Use fields in conjunction with field-activated (e.g. ferromag-netic) particles.

� Inventive Principle 13: ‘‘The other way round’’(a) Invert the action(s) used to solve the problem (e.g. instead

of cooling an object, heat it).(b) Make movable parts (or the external environment) fixed,

and fixed parts movable).(c) Turn the object (or process) ‘‘upside down’’.� Inventive Principle 12: Equipotentiality

(a) In a potential field, limit position changes (e.g. change oper-ating conditions to eliminate the need to raise or lowerobjects in a gravity field).

According to the specific characteristics of the diving fin, Inven-tive Principle 13 is selected in this case.

6.2.3. Heuristic abstract solution from inventive standardsIn this phase, the user is guided to work with semantic inven-

tive standards based on the chosen inventive principle.As shown in Fig. 12, the chosen inventive principle is firstly dis-

played, and the user needs to choose its most similar inventivestandard for the case of the ‘‘diving fin’’. The methods proposedin Section 5.1.2 can facilitate this process, based on which 5 mostsimilar inventive standards are returned. In this case, 5 most sim-ilar inventive standards for Inventive Principle 13 are:

� Inventive Standard 1.1.6: Minimum mode of action.� Inventive Standard 1.1.7: Maximum mode of action.


http://www.time-to-innovate.com/steps_matrix


Fig. 12. Semantic inventive standards.

Load the inven�ve standards ontology

Load triples

Process classes, proper�es and instances

(a) Loading the inventive standards ontology

Register the Rule engine

Execute the inference and build a new ontology to store the asserted and inferred result

The new ontology is iden�fied as: Id_A�erInven�veStandardOnto.owl

(b) Ontology reasoning

Fig. 13. The ontology inference for the heuristic abstract solutions.


Please cite this article in press as: W. Yan et al., IngeniousTRIZ: An automatic ontology-based system for solving inventive problems, Knowl. Based Syst.(2014), http://dx.doi.org/10.1016/j.knosys.2014.11.015


Fig. 15. Choose a way to improve the problem model.

Fig. 14. The obtained heuristic abstract solutions.


� Inventive Standard 4.3.3: Using resonance oscillation of attachedobject.� Inventive Standard 5.1.1.2: Bypass ways 2.� Inventive Standard 5.1.1.5: Bypass ways 5.

In order to implement Su-Field analysis, the user needs tochoose one from the obtained 5 similar inventive standards. In thiscase, Inventive Standard 5.1.1.5 is selected.

With the help of the chosen inventive standard, the user isguided to implement Su-Field analysis. Firstly, the user needs toprovide the basic information about the problem model, includingtwo substances and a field. This information will be used to instan-tiate the inventive standards ontology in the process of ontologyreasoning. Then, several related types of transformation areobtained based on the chosen inventive standards, from whichthe user needs to choose appropriate one for the specific case.There are two ways of representing the transformation, that is,the transformation of Su-Field models and the text description.


In this case, only one type of transformation - ‘‘A normal additiveis to be introduced in very small quantities and concentrated incertain parts of the object.’’ is obtained and used for the followingontology inference.

As presented in Fig. 13(a) and (b), ontology inference is exe-cuted to generate heuristic abstract solutions, which are stored ina new-built ontology. As shown in Fig. 14, for the case of the ‘‘div-ing fin’’, the elements (substances and field) in the solution modelare generated after ontology inference. At the same time, the heu-ristic abstract solutions are created automatically in terms of twoaspects:

� The modification in Substance1 – ‘‘water’’: A normal additive invery small quantities and concentrated in certain parts of‘‘water’’ is to be introduced.� The modification in Substance2 – ‘‘diving fin’’: A normal addi-

tive in very small quantities and concentrated in certain partsof ‘‘diving fin’’ is to be introduced.

Assuming that water cannot be modified in this case, we needto modify the diving fin by bringing in some additives. There areseveral kinds of modifications of the diving fin, for example, thechange of its structure and its shape by using liquid or gas. In orderto modify the diving fin, the user can continue to work with theheuristic physical effects to instantiate the solution model in reallife.

6.2.4. The search of heuristic physical effectsIn this phase, four ways to improve the problem model are

given, that is, ‘‘Add a Substance’’, ‘‘Modify a Substance’’, ‘‘Add aField’’ and ‘‘Modify a Field’’, and the user needs to select appropri-ate one from them. According to the inventive standard chosenabove, a heuristic way is indicated to facilitate this process. For thiscase, as presented in Fig. 15, ‘‘Modify a Substance’’ is suggested tomodify the problem model.

Then, in order to implement the ontology reasoning, the useralso needs to provide the level of granularity for the element (sub-stance or field) to be modified. Fig. 16 shows the level of granular-ity for substance, and in this case, ‘‘Solid’’ is chosen.

Finally, based on the information obtained above, the ontologyinference with the rules (ISRule 3, 4 and PERule 3, 4), presented in



Fig. 16. Choose the level of granularity for the substance to be modified.


Fig. 17(a), is executed to generate the heuristic physical effects,which are stored in a new-built ontology. The user can select anappropriate physical effect for solving the specific problem. Asshown in Fig. 17(b), for the case of the ‘‘diving fin’’, two physicaleffects – ‘‘Thixotropy’’ and ‘‘Acoustic Vibrations’’ are obtained,17

and their detailed description is displayed in the text box withimages to show the functional process.

Assumed that the effect ‘‘Thixotropy’’ is chosen to design theheuristic concept solution. ‘‘Thixotropy’’ is the property exhibitedby some gels or fluids that are generally viscous or thick under nor-mal conditions, but turn to a less viscous state when shaken, stir-red or agitated. These gels later take a certain period of time toreturn to their original state when allowed to stand without beingdisturbed. There exist many kinds of thixotropic fluids, such as,pseudoplastic fluids, natural fluids and biological fluids.

In this case, a shear thickening liquid can be used to modify thediving fin. On the one hand, when the divers push the diving fin,the shear thickening liquid changes to a solid state, making kickingefficiently, and on the other hand, the shear thickening liquidchanges back to a liquid state without buffering external force,which makes the diving fin lightweight, flexible and easy to use.

Taking this into account, the concept of tubular shear thicken-ing fins is proposed as shown in Fig. 18. This concept is a trulyinventive concept since there are no-existing widely distributedand produced products in industry based on this principle (onlyprototypes, patented fabrics, dampers in the truck industry, etc.).

7. Discussion

Although this automatic ontology-based system for solvinginventive problems can facilitate the process of using TRIZ effec-tively, several problems still need to be solved in the furtherresearch.

At first, through the case study presented in Section 6.2, theprogressiveness and appropriateness of the system for providingmore relevant and heuristic knowledge for helping TRIZ users toinnovate have been illustrated. However, the process of using TRIZto solve inventive problems is complex and several simple casesstudies are not enough to completely evaluate and validate the

17 In this application, 20 physical effects are used to test the related functions of theontology inference.


applicability and efficiency of this system. In order to demonstratethe applicability and usefulness of our approaches and the proto-type, more applications in real industrial cases are necessary andimportant. As a result, all the proposed methods in this researchwill be integrated in the research version of the software STEPSfor the industrial evaluation.18

Another problematic factor is WordNet, which is the semanticdictionary used in the calculation of semantic similarity of shorttexts. The proposed methods will not always be able to satisfythe requirements of some specialized problem because of the lim-its of the WordNet dictionary. The interaction with users will solvethis problem to a certain extent. For the moment, there is no strat-egy to lead the use of the SWRL rules to take profit of the knowl-edge about the specific case being solved. We intend, therefore,to provide a communication platform with users thanks to thedevelopment of SWRL rules describing the process of theinteraction.

Then, during the process of searching heuristic abstract solu-tions, TRIZ users start solving inventive problems with 40 inventiveprinciples to obtain an abstract solution, and then, according to theselected inventive principles, the similar items of 76 inventivestandards are obtained based on the semantic similarity calculatedin advance. With the help of these similar items, the useful heuris-tic abstract solutions are obtained through the ontology inference.However, the direct transformation from solving technical con-tradictions to implementing Su-Field analysis, from the resolutionwith 40 inventive principles to that with 76 inventive standards,results in a certain amount of missing information in the heuristicabstract solutions. For example, in the case of the ‘‘diving fin’’, onlythe specific parameters ‘‘kicking efficiency’’ and ‘‘ease of use’’ areconsidered to solve the technical contradiction, while the physicalparameters of diving fins or using diving fins, such as the circum-stance of its use and its components, are not considered, whichoften makes the user confused to modify the diving fin accordingto the obtained concept solution. One way to solve this problemis to bring in the analysis with 11 separation methods and physicalcontradictions to provide complementary information, that is, forthe chosen inventive principles, not only the similar inventive

18 Up to now, the proposed methods about automatic matching between SPs andGEPs, and searching heuristic abstract solutions from inventive standards have beencompletely integrated in the research version of STEPS.



Execute the ontology inference and build a new ontology to store the results

The new ontology is iden�fied as: Id_A�erPhysicalEffectOnto.owl

Load the physical effects ontology

Load triples

Process classes, proper�es and instances

(a) The ontology inference based on the physical effects ontology

List of heuris�c physical effects

Descrip�on for each physical effect

Figure showing the process of using each physical effect

(b) The detailed information for the heuristic physical effects

Fig. 17. The search of heuristic physical effects.


standards are used to generate the heuristic abstract solutions, butalso the similar separation methods are used to provide the infor-mation in physical level.

Last but not least, in order to solve various specific problems,the number and the content of the physical effects need to changedynamically according to the development of different kinds offields, for example, Visual Effects recently becoming accessibleowing to the appearance of the affordable animation and compos-iting software. The list of physical effects used in our research hasbeen proposed several years ago, and in order to keep its dynamic-ity, we intend to use text mining techniques to extract the useful


information for the physical effects from online resources – Wiki-pedia, such as, their initial and final states, and then instantiatethem automatically through Protege-OWL API.

8. Conclusions

In order to facilitate the resolution of inventive problems withTRIZ, an automatic system of solving inventive problems is pro-posed based on semantic similarity, ontology modeling and ontol-ogy inference. In this system, the TRIZ users start solving aninventive problem with the TRIZ knowledge source of their choice



Fig. 18. The heuristic concept solution of tubular shear thickening fins.


to obtain an abstract solution. According to the selected items ofthat first knowledge source, the similar items of other knowledgesources are obtained based on the semantic similarity calculatedin advance. With the help of these similar items and the heuristicphysical effects, other specific solutions are returned throughontology inference.

The contribution of the system IngeniousTRIZ can be summa-rized into three aspects:

Firstly, the TRIZ knowledge sources and physical effects are for-malized based on ontology modeling. Through the detailed analy-sis for the process of using the TRIZ knowledge sources andphysical effects, the related concepts and their relations aredefined in the ontologies. In comparison to traditional commercialsoftware, the formalization based on ontologies provides concep-tual resources for knowledge based systems (KBS) and makes itpossible to automate the process of solving inventive problemsby using ontology reasoning. It also permits the tracking of differ-ent applications to study and compare them based on ontologies,and, in this way, the improvement of the whole methodology.

Secondly, different methods for calculating semantic similarityare proposed to facilitate the process of solving inventive prob-lems. On the one hand, in order to search appropriate inventiveprinciples in the contradiction matrix, the match between specificparameters and generic engineering parameters is implementedautomatically based on semantic similarity, through which theefficiency and accuracy of choosing inventive principles to solveproblems have been improved greatly. On the other hand, themissing links among the TRIZ knowledge sources are defined basedon semantic similarity. Compared with the manual work, it usessemantic methods to compare the different abstract models orsolutions and gives an additional analysis to eliminate the variousinterpretations of different TRIZ users.

Finally, this system uses a method for solving inventive prob-lems with TRIZ in a different way from the existing approacheswidely used in the domain of inventive design, since our researchmakes it possible to transform the solution obtained with oneknowledge source to other that is obtained with another one thatwould not have been used by the TRIZ user, and therefore providesnew options to solve inventive problems with diversified knowl-edge sources. By using this system, on the one hand, more solu-tions for specific cases from different TRIZ knowledge sourcescan be obtained automatically, and so the efficiency of operatinginventive practice in design and also in the Research and Develop-ment (R&D) department can be improved. On the other hand, the


resolution with more than one TRIZ knowledge sources insteadof a single source can yield to more stable and appropriate specificsolutions, and so the risk of generating useless solutions could bereduced greatly.

Acknowledgement

This work is supported by the National Natural Science Founda-tion of China (No. 61272094) and Shandong Provincial Key Labora-tory Project.

References

[1] S. Agreste, P.D. Meo, E. Ferrara, D. Ursino, Xml matchers: approaches andchallenges, Knowl.-Based Syst. 66 (2014) 190–209.

[2] G. Altshuller, Creativity as An Exact Science, Gordon and Breach ScientificPublishers, New York, 1984.

[3] G. Altshuller, TRIZ The Innovation Algorithm; Systematic Innovation andTechnical Creativity, Technical Innovation Center Inc., Worcester, MA, 1999.

[4] K. Amailef, J. Lu, Ontology-supported case-based reasoning approach forintelligent m-government emergency response services, Decis. Support Syst.55 (1) (2013) 79–97.

[5] E. Bonjour, J. Renaud, Pilotage des systeme de connaissances de competences:Comment definir les concepts principaux, in: Proceedings of ColloqueInterrnational de Génie Industriel (CIGI), 2005.

[6] D. Cavallucci, T. Eltzer, Parameter network as a means for driving problemsolving process, Int. J. Comput. Appl. Technol. 30 (12) (2007) 125–136.

[7] O. Corcho, M. Fernandez-Lopez, A. Gomez-Perez, Methodologies, tools andlanguages for building ontologies. Where is their meeting point?, Data KnowlEng. 46 (1) (2003) 41–64.

[8] C. Fellbaum, WordNet: An Electronic Lexical Database, MIT Press, Cambridge:MA, 1998.

[9] E. Friedman-Hill, Jess in Action, Manning Publications Co., Greenwich, 2003.[10] E. Gabrilovich, S. Markovitch, Computing semantic relatedness using

Wikipedia-based explicit semantic analysis, in: International JointConference on Artificial Intelligence(IJCAI), 2007, pp. 1606–1611.

[11] R. Gil, M. Martin-Bautista, A novel integrated knowledge support system basedon ontology learning: model specification and a case study, Knowl.-Based Syst.36 (2012) 340–352.

[12] N. Guarino, R. Poli, The role of ontology in the information technology, Int. J.Hum. Comput. Stud. 43 (5/6) (1995) 623–965.

[13] I. Horrocks, P. Patel-Schneider, S. Bechhofer, D. Tsarkov, Owl rules: a proposaland prototype implementation, Web Semantics 3 (2005) 23–40.

[14] D. Lin, An information-theoretic definition of similarity, in: Proceedings of the15th International Conference on Machine Learning, ICML ’98, 1998, pp. 296–304.

[15] T. Pedersen, S. Pakhomov, S. Patwardhan, C. Chute, Measures of semanticsimilarity and relatedness in the biomedical domain, J. Biomed. Inform. 40 (3)(2007) 288–299.

[16] M. Rodríguez-García, R. Valencia-García, F. García-Sánchez, J. Samper-Zapater,Ontology-based annotation and retrieval of services in the cloud, Knowl.-Based Syst. 56 (2014) 15–25.

[17] G. Salton, M. Lesk, Computer Evaluation of Indexing and Text Processing,Prentice, Hall, Inc., Englewood Cliffs, NJ, 1971.

[18] K. Sparck-Jones, A statistical interpretation of term specificity and itsapplication in retrieval, J. Doc. (1972) 11–21.

[19] W. Yan, C. Zanni-Merk, F. Rousselot, An application of semantic distancebetween short texts to inventive design, in: Proceedings of the InternationalConference on Knowledge Engineering and Ontology Development,KEOD2011, 2011a, pp. 261–266.

[20] W. Yan, C. Zanni-Merk, F. Rousselot, D. Cavallucci, A method of facilitatinginventive design based on semantic similarity and case-based reasoning, in:TRIZ Future 2011, 2011.

[21] W. Yan, C. Zanni-Merk, F. Rousselot, D. Cavallucci, P. Collet, A heuristic methodof using the pointers to physical effects in su-field analysis, in: TRIZ Future2012, 2012.

[22] C. Zanni-Merk, D. Cavallucci, F. Rousselot, An ontological basis for computeraided innovation, Comput. Ind. 60 (2009) 563–574.

[23] C. Zanni-Merk, D. Cavallucci, F. Rousselot, Use of formal ontologies as afoundation for inventive design studies, Comput. Ind. 62 (3) (2011) 323–336.


http://refhub.elsevier.com/S0950-7051(14)00415-8/h0005












































Documents

IngeniousTRIZ: An automatic ontology-based system for solving inventive problems