The semantic method for agents knowledge representation in the Cognitive Integrated Management Information System

Positio Papers of the Federated Coferece o Computer Sciece ad Iformatio Systems pp. 195 202 DOI: 10.15439/2015F319 ACSIS, Vol. 6 The sematic method for agets kowledge represetatio i the Cogitive Itegrated Maagemet Iformatio System Marci Heres Wrocław Uiversity of Ecoomics ul. Komadorska 118/120, 53-345 Wrocław, Polad Email: marci.heres@ue.wroc.pl Abstract This paper presets a method for agets kowledge represetatio by usig sematic etwork with ode ad liks activatio level defied o the istace, cocept, relatio ad axiom level. The first part shortly presets the state-of-the-art i the cosidered field; ext, the CIMIS prototype is shortly characterized; the formal defiitio of a method for agets kowledge represetatio is preseted i the last part of paper. C I. INTRODUCTION ONTEMPORARY the etire ecoomy is based o iformatio ad kowledge, therefore compaies must employ systems which support the kowledge maagemet process takig ito cosideratio the risk ad ucertaity of ecoomic decisios. Ofte the itegrated maagemet iformatio systems (IMIS) are used i this purpose. They are characterized by full itegratio both at the system/applicatio level ad the busiess process level. Note, however, that the properties of cotemporary IMIS are becomig more ad more iadequate. Apart from collectig ad aalyzig data ad geeratig kowledge, the system should also be able to uderstad the meaig of pheomea occurrig aroud the orgaizatio. It is becomig more ad more ecessary to make decisios based ot oly o kowledge but also o experiece, thus far regarded as purely huma domai [4]. I order to accomplish tasks set by IMIS, a multi-aget system ca be used cosist of several cogitive agets. Not oly do they eable quick access to iformatio ad quick search for the required iformatio, its aalysis ad coclusios, but also, besides beig resposive to eviromet stimuli, they have cogitive abilities that allow them to lear from empiric experiece gaied through immediate iteractio with their eviromets [15], which cosequetly allows a umber of decisio versios to be automatically geerated ad to make ad execute decisios. As a result of its ruig, the aget obtais kowledge of the eviromet i which it operates. If we desire to use this kowledge the it must be represeted i the form of a This work was fiacially supported by the Natioal Sciece Ceter (decisio No. DEC-2013/11/D/HS4/04096) specific structure. The ability of cogitive agets to uderstad the meaig of pheomea occurrig aroud the orgaizatio causes that the sematic structures have to be used. The otologies which are maily applied for this purpose iclude sematic etworks. They are formally defied i umerous papers (e.g. [9], [19], [27], [29]), however to a small extet they iclude the method for represetatio of the ucertaity of ecoomic decisios. Therefore, the sematic et with odes ad liks activatio level is a better solutio to represet the compaies kowledge. This type of represetatio eables processig both, kowledge represeted i a symbolic way, ad kowledge represeted i a umerical way. It also eables processig structured ad ustructured kowledge. Thus, it is possible to determie the certaity level of odes (cocepts) ad sematic relatios betwee these odes. The first suggestio to use such a structure, called slipet, is preseted i the Copycat project [14]. However, the existig papers lack the formal defiitio of this structure ad do ot cosider issues related to istaces ad relatios betwee odes ad istaces. This defiitio is however ecessary maily i order developed methods for the agets kowledge processig ad itegratio. The aim of this paper is to develop a formal defiitio of a method for agets kowledge represetatio usig sematic etwork with ode ad liks activatio level takig ito cosideratio the istace, cocept, relatio ad axiom (relatios betwee odes ad istaces) level. This structure has bee implemeted i the architecture of the cogitive agets ruig i the Cogitive Itegrated Maagemet Iformatio System (CIMIS) prototype. This paper is orgaized as follows: the first part shortly presets the state-of-the-art i the cosidered field; ext, the CIMIS prototype is shortly characterized; the formal defiitio of a method for agets kowledge represetatio is preseted i the last part of paper. II. RELATED WORKS I this sectio the methods related maily to cogitive agets kowledge represetatio will be aalyzed. I the study [6] cosiderig the taxoomy of cogitive aget c 2015, PTI 195

196 POSITION PAPERS OF THE FEDCSIS. ŁÓDŹ, 2015 architectures with respect to their kowledge represetatio ad learig methods, three mai groups of the architectures were distiguished: 1. Symbolic architectures which use declarative kowledge icluded i relatios recorded at the symbolic level, focusig o the use of this kowledge to solve problems. 2. Emerget architectures usig sigal flows through the etwork of umerous, mutually iteractig elemets, i which emerget coditios occur, possible to be iterpreted i a symbolic way. 3. Hybrid architectures which are the combiatios of the symbolic ad emerget approach, combied i various ways. The literature of subject presets may differet methods for agets kowledge represetatio used i the metioed groups. The mai of them iclude first-order predicate logic, productio systems, artificial eural etworks, frame represetatio, otologies such as sematic web, sematic etworks ad topic maps, multi-attributes ad multi-values structures, multi valued logic icludes a three valued logic ad a fuzzy logic. Some of these methods are closely related to the sematic agets kowledge represetatio. Productio systems are based o the productio rules cosistig of two elemets: cosequece or the head of the rule ad the other is the atecedece or the body of the rule which should be true to satisfy the cosequece part [16], [23]. Artificial eural etworks are geerally defied as systems of itercoected "euros" which ca compute values from iputs, ad are capable of machie learig as well as patter recogitio thaks to their adaptive ature [1], [18]. Actios take by the aget are directly coected with perceptio without the mediatio of symbolic reasoig. However, a area of applicatio is limited to agets coceptually simple recogizig actios or uderlyig actios directly related to the recogitio (for example, graspig objects). It ofte happes that agets kowledge is represeted as multi-attribute ad multi-value structure cosistig of differet umber of differet types of attributes. It allow for represetig the real word eviromet i a wide scope of objects features. Such structures are used, for example i case of the weather forecastig multiaget system [21] or supply chai maagemet multiaget system [13], [25]. Frame represetatio captures the way agets typically thik about Special Sectio of their kowledge, provide a cocise structural represetatio of useful relatios, ad support a cocise defiitio-by-specializatio techique that is easy for most domai experts to use. I additio, special purpose deductio algorithms have bee developed that exploit the structural characteristics of frames to rapidly perform a set of ifereces commoly eeded i kowledge-system applicatios. The taxoomic relatioships amog frames eable descriptive iformatio to be shared amog multiple frames (via iheritace) ad because the iteral structure of frames eables sematic itegrity costraits to be automatically maitaied. [8], [28]. Otology, i tur, most ofte is defied by the followig elemets [9], [10] O C, I, R, Z (1) where: C Set of cocepts (classes) I Set of istaces of cocepts R Set of biary relatios defied o C Z Set of axioms which are formulae of first-order logic ad ca be iterpreted as itegrity costraits or relatioships betwee istaces ad cocepts, ad which caot be expressed by the relatios i set R, or as relatioships betwee relatios icluded i R 1. The Sematic Web allows searchig ot oly iformatio but also kowledge. Its mai purpose is itroducig structure ad sematic cotet i the huge amout of ustructured or semi-structured distributed kowledge available o the Web, beig the cetral otio behid the Sematic Web that of otologies, which describe cocepts ad their relatios i a particular field of kowledge [27]. Sematic etworks [26] are kowledge represetatio schemes ivolvig odes ad liks (arcs or arrows) betwee odes. The odes represet objects or cocepts ad the liks represet relatios betwee odes. The liks are directed ad labeled; thus, a sematic etwork is a directed graph. I prit, the odes are usually represeted by circles or boxes ad the liks are draw as arrows betwee the circles. As a otological kowledge represetatio, the topic map stadard, itroduced by Iteratioal Orgaizatio for Stadardizatio (ISO/IEC 13250:2000), is also used. The topic maps are a kid of a sematic etwork, ad they allow writig iformatio of the data otology ad data taxoomy i a sematically ordered maer [17]. The topic map, most ofte, cosists of paret-child relatios. Preseted methods for kowledge represetatio ca be used oly i respect of the oe group of cogitive aget represetatio. For example, the topic maps ca be used i symbolic architectures ad eural etworks ca be used i emerget architectures. I the hybrid architectures ofte a combiatio of differet methods is used depedig o a task, which is to be executed. For example, if agets purpose is visual quality aalysis of grais, the eural etwork is used for recogizig the features of grais (e.g. shape, color) ad sematic etwork is used to determie the grai species. However, the disadvatages of this approach are the eed to implemet two (or more) differet types of modules for kowledge storig ad the complexity of the procedures for coversio of kowledge represeted usig eural etwork with the kowledge represeted by a 1 This defiitio differs from that of "axioms" i geerative grammar ad formal logic. I otology disciplies, axioms iclude oly statemets asserted as a priori kowledge.

MARCIN HERNES: THE SEMANTIC METHOD FOR AGENTS KNOWLEDGE REPRESENTATION 197 sematic etwork. Therefore, a better approach is to use methods that allow represetatio of both symbolic ad umerical kowledge i a itegrated, uiform maer. The first suggestio to use such a method, called slipet, is preseted i the Copycat project. This method is developed i the LIDA cogitive aget [7]. This hybrid architecture allows for symbolic ad emerget kowledge processig ad it uses the sematic et with ode ad liks activatio level (the slipet ) to represet kowledge. This type of represetatio eables processig kowledge represeted i a symbolic way, as well as kowledge represeted i a umerical way. Thus, it is possible to determie a certaity level of sematic relatios betwee odes (topics). Because the formal, mathematical defiitio of slipet has ot yet bee defied ad existig papers do ot cosider issues related to istaces ad relatios betwee odes ad istaces, it is very difficult to develop methods for advaced processig ad itegratio of kowledge represetig by this structure i CIMIS prototype, preseted i the ext part of this paper. III. THE COGNITIVE INTEGRATED MANAGEMENT INFORMATION SYSTEM I order to develop a prototype of CIMIS the followig subsystems have bee created (Fig 1.): fixed assets, logistics, maufacturig maagemet, huma resources maagemet, fiacial ad accoutig, cotrollig, busiess itelligece [2]. The fixed assets sub-system icludes support for the realizatio of processes related to fixed asset ad ivolved their depreciatio. The logistics sub-system has all the mai features supportig the employees of logistics departmet i their effective work [10]. The logistics sub-system eables maitaiig optimal stock to meet the eeds of productio departmet. The maufacturig maagemet sub-system support a processes related to a maufacturig executio. It iclude fuctios from the scope of the techical preparatio of productio capacity, productio plaig, material cosumptio plaig, plaig ad executio of a maufacturig tasks, maufacturig cotrol, visualizatio, moitorig ad archivig. The huma resources maagemet sub-system supports realizatio of such processes, as the employees of the compay data ad cotract registerig, recordig of workig time, wage calculatio, creatig the tax ad social security declaratio. The fiacial-accoutig sub-system supports registerig, to the full extet, ecoomic evets, also provides importat, from the poit of view of busiess maagemet, iformatio, cocerig, iter alia, paymet capacity, reveues, costs, fiacial result. Cotrollig sub-system is automatically processig data related to profit ad loss accout i cooperatio with accoutig sub-system. The cotrollig sub-system cosists of both a strategic ad operatioal cotrollig. The CRM sub-system is egaged i matters coected with esurig the best compay-customer relatios ad collectig iformatio i the customers' prefereces i terms of product purchase i order to icrease sales. The eterprise s eviromet moitorig is also realized by this sub-system. The purpose of busiess itelligece sub-system is to eable easy ad safe access to iformatio i a compay, operatio of its aalysis ad distributio of reports withi the compay ad amog its busiess parters, which i tur eables quick ad flexible decisio makig. I the cotext of, most of all, the busiess itelligece sub-system, but other sub-systems as well, the CIMIS makes cogitive visualizatio [36] features available, meaig it eables a visualizatio of multi-dimesioal data i oe picture that allows fidig the source of a problem i a short time ad cotributes to creatig ew kowledge about a object or problem [25]. Fig 1. The CIMIS architecture. Source: ow work.

198 POSITION PAPERS OF THE FEDCSIS. ŁÓDŹ, 2015 The system assumes that all agets are at 'ot-taught' status i the iitial phase. They ca be iitially grouped accordig compay's eeds for sub-systems. For i-stace, oe group of agets is assiged to the logistics sub-system, aother oe is assiged to the maufacturig maagemet sub-system ad yet aother oe to fiacial ad accoutig sub-system. Withi the groups, the agets ca be iitially 'taught' by the compay that implemets the system. Next stages of learig for both grouped ad ugrouped agets are doe by the compay staff. Agets ca also lear without teacher through aalyzig the results of their decisios. The agets of all sub-systems cooperate themselves i order to better busiess processes realizatio. For example, the eterprise s eviromet moitorig results performed by CRM sub-system aget are usig by the other agets. The mai operatig purpose of the Supervisor aget is to moitor the proper operatio of other agets, maily i the field of detectio ad solvig coflicts of kowledge ad experiece. The aget aalyzes, i close-to-real time, the structures of kowledge ad experiece of all agets. Wheever a coflict occurs, it employs a solutio algorithm based o a method that uses cosesus theory [9, 21], ad the result of the aget's actios is accepted by the system as curret state of kowledge ad experiece. Note that all CIMIS sub-systems are coected by a sigle, coheret stream of i-formatio ad kowledge available olie to the maagemet, because owadays attetio is paid to fuctioal complexity, maagig all fields of operatio i a compay, proper flow of iformatio ad kowledge amog sub-systems as well as the ability to perform a variety of aalyses ad to create reports for maagemet. The implemetatio of this solutio is realized as follow: 1. Commuicatio betwee modules of agets architecture was esured by usig LIDA framework s codelets, 2. Commuicatio betwee agets is based o Java Message Service (JMS) techology. The represetatio of iformatio ad kowledge (geerated i result of agets operatig) i form of XML format documet, was adopted (the JMS messagig is at the text type). The commuicatio is realized i publish/subscribe messagig domais it guaratees that iformatio or kowledge geerated by oe of agets is immediately available for the other agets. The asychroous message cosumptio is used. All of the sub-systems fuctios are available as a local services or e-services (e.g. e-busiess, e-procuremet, e- paymet) by usig Web Services techology. At the physical level, the IMIS is built o the basis of the mai two techologies the LIDA framework (due to framework is developed at Java laguage ad it is ope the implemetatio of the other Java techologies metioed JMS, Java Database Coectivity or Java API for XML Web Services - is possible) ad Microsoft SQL Server 2008 database maagemet system. The realizatio of the CIMIS is based o the LIDA cogitive aget architecture [7, 24], which is of emergetsymbolic ature, owig to which the processig of both structured ad ustructured kowledge is possible. I the LIDA architecture it was adopted that the majority of basic operatios are performed by the so-called codelets, amely specialized, mobile programs processig iformatio i the model of global workspace. The fuctioig of the cogitive aget is performed withi the framework of the cogitive cycle ad it is divided ito three phases: the uderstadig phase, the cosciousess phase ad the selectio of actios ad learig phase. At the begiig of the uderstadig phase the stimuli received from the eviromet activate the codelets of the low level features i the sesory memory [4]. The outlets of these codelets activate the perceptual memory, where high level feature codelets supply more abstract thigs such as objects, categories, actios or evets. The perceptio results are trasferred to workspace ad o the basis of episodic ad declarative memory local liks are created ad the, with the use of the occurreces of perceptual memory, a curret situatioal model is geerated; it other words the aget uderstads what pheomea are occurrig i the eviromet of the orgaizatio. The cosciousess phase starts with formig of the coalitio of the most sigificat elemets of the situatioal model, which the compete for attetio so the place i the workspace, by usig attetioal codelets. The cotets of the workspace module are the trasferred to the global workspace (the so-called broadcastig takes place), simultaeously iitializig the phase of actio selectio. At this phase possible actio schemes are take from procedural memory ad set to the actio selectio module, where there compete for the selectio i a give cycle. The selected actios activate sesory-motor memory for the purpose of creatig a appropriate algorithm of their performace, which is the fial stage of the cogitive cycle [3]. The cogitive cycle is repeated with the frequecy of 5 10 times per secod. Parallelly with the previous actios the aget s learig is performed, which is divided ito perceptual learig cocerig the recogitio of ew objects, categories, relatios; episodic learig which meas rememberig specific evets: what, where, whe, occurrig i the workig memory ad thus available i the awareess; procedural learig, amely learig ew actios ad actio sequeces eeded for solvig the problems set; coscious learig relates to learig ew, coscious behaviors or stregtheig the existig coscious behaviors, which occurs whe a give elemet of the situatioal model is ofte i the workspace. The aget s learig may be performed as learig with or without a teacher. It is worth emphasizig that each cogitive aget supportig decisio-makig must have the ability of

MARCIN HERNES: THE SEMANTIC METHOD FOR AGENTS KNOWLEDGE REPRESENTATION 199 groudig the symbols, amely assig relevat real world objects to specific symbols of the atural laguage. This is ecessary to correctly process ustructured kowledge saved maily by meas of the atural laguage ad thus, for istace, the cliets opiios o products. The kowledge of this type is curretly becomig more ad more sigificat for a compay because it may have impact o its competitiveess level. For istace aalyzig the cliets opiios o a give product, the sales volume of a give product i the future may be estimated (of course with a certai level of probability). The Cogitive Computig Research Group established by S. Frakli, developed i 2011 the framework (i Java laguage) sigificatly facilitatig the implemetatio of the cogitive aget i CIMIS. It should also be emphasized that the whole framework code is ope, i.e. the developer has access to the defiitios of all methods. The learig mechaisms are ot implemeted i curret versio of the framework (they are uder implemetatio by CCRG). This framework, however, does ot cotai the mechaism for automatically storig the aget s kowledgebase i a physical database. After the power is tured off, the agets kowledgebase is lost. Iitially the storig of a agets kowledgebase i database has bee lauched after realizig by give codelet its task. However, this method proved to be isufficiet i case more complex tasks. Therefore the eed appears for developig a method for permaet, automatically storig the aget s kowledgebase i a physical database. The ext part of article presets a formal, mathematical defiitio of a sematic method for agets kowledge represetatio based o sematic et with ode ad liks activatio level which icludes istaces ad relatios betwee odes ad istaces. IV. THE METHOD FOR AGENTS KNOWLEDGE REPRESENTATION The mai ideas of usig sematic etwork as agets kowledge represetatio are as follows [5]: the meaig of a symbol or cocept stems from relatioships with other symbols ad cocepts; the huma memory is a etwork of associatios, iformatio is cotaied i the odes ad arcs (liks) coectig the odes (ode = cocept; i the brai it is a patter of beats activity of may euros), every cocept is a etwork ode, the liks betwee odes are clearly preseted, the liks ca be of differet types, the sematic etwork is a model of episodic memory, but also sematic memory, the odes represet, amog other thigs: objects, types, or classes, evets, activities, episodes, places, times, liks represet, amog other thigs: to give a example, subclass, the is-a relatioship, it is part of somethig, logical cojuctios ad, or, actios, istrumets. The eed for agets kowledge represetatio by sematic et with ode ad liks activatio level results maily from the followig presumptios: i a huma brai from phoology ad graphemes of the word to its meaig ad model of the situatio, we have differet patters of distributio (levels) of stimulatio (activatio), ad associatios betwee them [5], Pulvermuller [22] states, that because sematic activatio followed by 90 ms the phoological activatio the a brai stimulatio is a atural base of sematic represetatio; very importat issue is probability distributio (activatio level); cocepts related to the same topic better fit together ad create a coheret cocept graph of a active part of sematic memory icludig the ihibitio of the activatio ad propagatio. together with odes ad liks ad their activatio level the istaces ad axioms have to be icluded i the sematic etwork; it facilitate a automatic storig a agets kowledgebase i the physical database. Takig ito cosideratio: preseted presumptios, the defiitio of otology preseted i sectio 2, ad a slipet features, the method for agets kowledge represetatio, called slipetplus is defied as follows: Defiitio 1 The slipetplus is called a quadruple: SN N, I, L, Z (2) where: N set of odes, I set of istaces of odes, L set of liks i.e. set of fuzzy relatios defied o N, Z set of axioms. This defiitio exteds the slipet preseted by [14, 24] about the set of istaces ad the set of axioms. Let us to defie the particular elemets of slipetplus. We assume a real world <O, V> where O is a fiite set of objects ad V is the domai of O; that is, V is a set of object values, ad V o O V (3) o where V o is the domai of object o. We cosider the slipetplus referrig to the real world (O, V) - such slipetplus is called <O, V>-based. The slipetplus detailed defiitios must be cosidered o the four levels: the ode level. the istace level. the lik level. the axiom level. These defiitios are developed i the ext subsectios of this paper.

200 POSITION PAPERS OF THE FEDCSIS. ŁÓDŹ, 2015 A. The ode level Defiitio 2. A ode of a <O, V>-based slipetplus is defied as a triple: Node, O, V (4) where is the uique ame of the ode, O c O [0,1 ] is a fuzzy set of objects represeted by ode with a certai level of probability, ad V c V [0,1 ] is the objects domai: Nested pair O, V c V V [0,1] (5) o, v O o is called the structure of ode. It is obvious that all odes belogig to the same slipetplus are differet from each other. However, otice that withi a slipetplus there may be two or more odes with the same structure. Such a situatio may take place, for example, for odes perso ad body. For expressig the relatioship betwee them the liks from set L will be very useful. Set N i the slipetplus defiitio is a set of odes ames ad their activatio levels. B. The istace level Defiitio 3 A istace of a ode is described by the objects from set O with values from set V ad is defied as a pair: istace i, v (6) where i is the uique idetifier of the istace i world <O, V> ad v is the value of the istace a tuple of type O, ad ca be preseted as a fuctio: v : O [0,1] V [0,1] (7) such that v( o, p) V for all o o p O,. Value v is also called a descriptio of the istace withi a object. A ode may be iterpreted as a set of all istaces described by its structure. We ca the write i for presetig the fact that i is a istace of ode. All istaces of the same odes withi a slipetplus should be differet from each other. The same istace may belog to differet odes ad may have differet values. C. The lik level I a slipetplus withi a pair of odes there may be defied oe or more liks. Liks betwee odes describe the relatioships betwee them. For example, betwee two odes may be defied such relatios as Syoym relatio or Atoym relatio. Liks betwee odes are icluded i set L of the slipetplus defiitio. Defiitio 4. Let set N of odes is give. The lik is called the followig relatio: L : N N [0,1] (8) i a space N N. D. The axiom level The set Z of axioms are formulae of fuzzy logic ad ca be iterpreted as itegrity costraits or relatioships betwee istaces ad odes, ad which caot be expressed by the relatios i set L. Defiitio 5. Let set N of odes ad set I of istaces are give. The axiom is called the followig relatio: Z : N I [0,1] (9) i a space N I. E. A graphical represetatio ad a implemetatio of the slipetplus The developed defiitio of slipetplus ca be visualized i a graphical form. The Fig 2 presets a example of a graphical represetatio of the slipetplus. Fig 2. The example of a graphical represetatio of the slipetplus. Source: ow work. The arrows draw with a cotiuous lie represet liks, while the arrows draw dotted lies deote the axioms. The preseted slipetplus cosist of three odes with their activatio levels. The iterpretatio of the Node1 is as follows: The Node1 exists i the real world with a probability level 0.75. Iterpretatio of other odes is similar. The Node1 is coected with Node2 ad Node3 by liks with the level of probability respectively 0.43 ad 0.48. The iterpretatio of the Istace1 is as follows: The Istace1 exists i the real world with a probability level 0.55. Iterpretatio of other istaces is similar. The Node1 is coected with Istace1 by axiom with the level of probability 0.55. The Node1 is coected with Istace3 by axiom with the level of probability 0.75 ad Node3 is also coected with the Istace3 but with the level of probability 1.0. The slipetplus preseted o Fig 2. ca represet, for example, followig practical ecoomic situatio: Let: Node1 deotes a Ivestmet, Node2 deotes a Securities, Node2 deotes a Currecies Istace1 deotes a Gold, Istace2 deotes a Estate,

MARCIN HERNES: THE SEMANTIC METHOD FOR AGENTS KNOWLEDGE REPRESENTATION 201 Istace3 deotes EUR, Istace4 deotes Compay1, Istace5 deotes Compay2. I cosidered situatio it is ecessary to ivest o the 0.75 probability level 2. The accessibility of Securities equals 0.34 3 ad accessibility of Currecies equals 0.25. It is better to ivest i Currecies (probability level 0.68) tha i Securities (probability level 0.43). The accessibility of: Gold equals 0.55, Estate equals 0.75, EUR equals 0.25, Compay1 security equals 0.15 ad Compay2 security equals 0.75. The axioms deote that the probability to achieve a positive (satisfactory) rate of retur i case ivestmet i the Gold equals 0.56 ad i case securities of Compay2 equals 0.2. The iterpretatio of remaiig axioms is similar. Takig ito cosideratio the implemetatio issues, the slipetplus has bee implemeted i CIMIS. The code of LIDA framework classes related to slipet implemetatio has bee exteded to the slipetplus implemetatio. The LIDA agets kowledgebase is automatically stored i a database by usig istaces. The object-orieted osql database model is suitable for storig odes, liks, istaces, axioms together with their activatio levels. V. CONCLUSION The formal, mathematical method for agets kowledge represetatio uses sematic et with ode ad liks activatio level takig ito cosideratio istaces ad relatios betwee odes ad istaces, has bee developed i this paper. This method has bee implemeted i CIMIS ad ca be directly implemeted i other multiaget systems. It should be oted, that this type of represetatio allows processig both kowledge represeted i a symbolic way, ad kowledge represeted i a umerical way. Simultaeously a structured ad ustructured kowledge ca be processed. Thus, it is possible to determie a certaity level of sematic relatios betwee odes (topics). I case of the ecoomic kowledge, it is a very importat issue because the decisios-makig process based o the type of kowledge usually takes place i coditios of risk ad ucertaity. The implemetatio of the developed sematic method for agets kowledge represetatio greatly facilitated the agets kowledge base mappig i the physical database. The developed method eables the realizatio of further research works related to developig methods for processig ad itegratio kowledge represeted by the slipetplus. 2 A probability level is calculated, for example, o the basis of historical rates of returs ad risk level (e.g. by usig Value at Risk ad Sharpe ratio measures). 3 For example, we wat to ivest 1000 EUR, but at the price we ca pay up oly securities i the amout of 340 EUR are available other securities are too expesive i order to achieve the positive rate of retur. Also actio selectio ad actio performig procedures are uder implemetatio. REFERENCES [1] O. Badawy ad A. Almotwaly, Combiig eural etwork kowledge i a mobile collaboratig multi-aget system, Electrical, Electroic ad Computer Egieerig,. ICEEC '04, Iteratioal Coferece o,2004, pp.325,328, doi: 10.1109/ICEEC.2004.1374457 [2] A. Bytiewski (ed.), Architecture of itegrated maagemet iformatio systems, Wroclaw Uiversity of Ecoomics Press, Wroclaw 2005. [3] A. Bytiewski, A. Chojacka-Komorowska, M. Heres ad K. Matouk, The Implemetatio of the Perceptual Memory of Cogitive Agets i Itegrated Maagemet Iformatio System, i: D. Barbucha, N. T. Nguye, J. Batubara, New Treds i Itelliget Iformatio ad Database Systems, Studies i Computatioal Itelligece Volume 598, Spriger Iteratioal Publishig Switzerlad, 2015, pp 281-290. doi: 10.1007/978-3-319-16211-9_29 [4] Cogitive Computig Research Group, http://ccrg.cs.memphis.edu/, data odczytu: 29.10.2014. [5] W. Duch, Sztucza Iteligecja Reprezetacja wiedzy II: sieci sematycze, https://www.fizyka.umk.pl/~duch/wyklady/ai/ai06-1.ppt [02.05.2015]. [6] W. Duch, Architektury kogitywe, czyli jak zbudować sztuczy umysł, i: R. Tadeusiewicz (ed.) Neurocyberetyka teoretycza, Wydawictwa Uiwersytetu Warszawskiego, Warszawa 2010. [7] S. Frakli, F. G. Patterso, The LIDA architecture: Addig ew modes of learig to a itelliget, autoomous, software aget, i: Proc. of the It. Cof. o Itegrated Desig ad Process Techology, Sa Diego, CA: Society for Desig ad Process Sciece, 2006. [8] R. Fikes ad T. Kehler., The role of frame-based represetatio i reasoig. Commu. ACM 28(9), 1985, pp. 904-920. DOI=10.1145/4284.4285. [9] D. Fesel, Otologies: Silver Bullet for Kowledge Maagemet ad Electroic Commerce, Spriger-Verlag, New York, 2001. [10] T.R. Gruber, A Traslatio Approach to Portable Otology Specificatios, Kowledge System Laboratory, Academic Press Staford Uiversity, 1993. [11] M. Heres, A Cogitive Itegrated Maagemet Support System for eterprises, i D. Hwag, J. Jug, N.T. Nguye (eds.), Computatioal Collective Itelligece Techologies ad Applicatios, Lecture Notes i Artificial Itelligece, vol. 8733, Spriger-Verlag, 2014, pp. 252-261. doi: 10.1007/978-3-319-11289- 3_26 [12] M. Heres ad N.T. Nguye, Derivig Cosesus for Hierarchical Icomplete Ordered Partitios ad Coverigs, Joural of Uiversal Computer Sciece 13(2)/2007, pp. 317-328. [13] M. Heres ad J. Sobieska-Karpińska, Applicatio of the cosesus method i a multiaget fiacial decisio support system, Iformatio Systems ad e-busiess Maagemet, Spriger Berli Heidelberg 2015, doi: 10.1007/s10257-015-0280-9. [14] D. Hofstadter ad M. Mitchell, The copycat project: A model of metal fluidity ad aalogy-makig, i D. Hofstadter ad the Fluid Aalogies Research group, Fluid Cocepts ad Creative Aalogies. Basic Books. Chapter 5, 1995. [15] R. Katarzyiak, Grutowaie modalego języka komuikacji w systemach agetowych, Akademicka Oficya Wydawicza EXIT, 2007. [16] M. A. Kadhim, A. Alam ad M. K. Harlee, A Multi-itelliget Aget Architecture for Kowledge Extractio: Novel Approaches for Automatic Productio Rules Extractio, Iteratioal Joural of Multimedia & Ubiquitous Egieerig; Vol. 9 Issue 2, 2014, p.95. [17] J. Korczak, H. Dudycz ad M. Dyczkowski, Desig of Fiacial Kowledge i Dashboard for SME Maagers, i: Proceedigs of the 2013 Federated Coferece o Computer Sciece ad Iformatio Systems, 2013, pp. 1111-1118. [18] J. Korczak ad F. Hammadi-Mesmoudi, A way to improve a architecture of eural etwork classifier for remote sesig applicatio, Neural Processig Letters 1(1), pp.13-16, 1994.

202 POSITION PAPERS OF THE FEDCSIS. ŁÓDŹ, 2015 [19] S. P. Li, Q. W. Yi, Y. J. Hu et al., Overview of researches o otology Joural of Computer Research ad Developmet, 2004, 41(7), pp.1041-1052. [20] M. Maleszka ad N.T. Nguye, Itegratio computig ad collective itelligece, Expert Systems with Applicatios, vol. 42 (1),, 2015, pp. 358-378. doi:10.1016/j.eswa.2014.07.036 [21] N. T. Nguye, Processig icosistecy of kowledge i determiig kowledge of collective, Cyberetics ad Systems: A Iteratioal Joural, 40 (8), 2009, pp. 670-688. [22] F. Pulvermuller, The Neurosciece of Laguage. O Brai Circuits of Words ad Serial Order, Cambridge Uiversity Press 2003. [23] X. Z. Wag, S. F. A., Research o learig weights of fuzzy productio rules based o maximum fuzzy etropy, Joural of Computer Research ad Developmet, 43(4),2006, pp.673-678. doi: 10.1360/crad20060416 [24] J. Saider, R. McCall ad S. Frakli, The LIDA Framework as a Geeral Tool for AGI The Fourth Coferece o Artificial Geeral Itelligece, 2011. doi: 10.1007/978-3-642-22887-2_14 [25] J. Sobieska-Karpińska ad M. Heres, Cosesus determiig algorithm i multiaget decisio support system with takig ito cosideratio improvig aget's kowledge, i: Proceedigs of the 2013 Federated Coferece o Computer Sciece ad Iformatio Systems, Wrocław 2012, pp. 1035-1040. [26] J. F. Sowa, Sematic Networks, http://www.jfsowa.com/pubs/ semet.htm [02.05.2014]. [27] Z. Zeg, Costructio of kowledge service system based o sematic web, Joural of The Chia Society For Scietific ad Techical Iformatio, 24(3), 2005, pp.336-340. [28] G. J. Zhu ad Y. M. Xia, Research ad practice of frame kowledge represetatio, Joural of Yua Uiversity (Natural Scieces Editio), 28(S1), 2006, pp.154-157. [29] T. Ataasova, T., Towards sematic-based process-orieted cotrol i digital home, Federated Coferece o Computer Sciece ad Iformatio Systems (FedCSIS), 2014, pp.1133-1137, doi: 10.15439/2014F317.