0HGLDWLQJ,GHDVLQDQ$JHQWEDVHG7HDPIRU%XVLQHVV 3URFHVV5HHQJLQHHULQJ7RZDUGD/LQJXLVWLF2QWRORJ\ $OH[DQGUD*DODWHVFX 1DWLRQDO ,QVWLWXWH IRU 5 ' LQ ,QIRUPDWLFV  $YHUHVFX $YHQXH  %XFKDUHVW  520$1,$ )D[      (PDLODJDO#VWGLFLUR ABSTRACT 2. TYPES AND ROLES OF THE This statement first outlines the main requirements for the ONTOLOGIES FOR BPR application of the ontologies and of the agent-based technology to The ontology-based comunication among the members of the Business Process Reengineering (BPR). Then it tries to motivate BPR team is needed because they usually have different the need for a new type of mediation for BPR, namely the specializations and approaches on the target processes and they mediation of ideas (instead of services, as the existing facilitators need to arrive at a common understanding of the concepts, provide), in order to automatically collect, compare, combine, actions, solutions etc and, eventually, at a common language to analyze ideas and, then, infer upon them, during the brainstorming express their ideas. meetings. The statement also presents, as a possible solution to the ideas mediation and ontology integration, the intended steps in The communication between the software agents for BPR is the construction of an upper-level ontology with linguistic supposed to rely on three ontologies: (1) domain ontology and features. (2) BPR ontology, for the content of the messages; (3) the communication ontology, for the communication protocol. These ontologies will have the following roles in BPR: 1. INTRODUCTION 1. Domain ontology (e.g. for manufacturing, education, Business Process Reengineering (BPR) means the rethinking and insurance, etc) ascribes the meaning to the concepts/ symbols redesign of the business processes, mainly by the analysis and in the agents' messages regarding the target domain. In [2], design of the team-based work flows and processes within and the brainstorming manual implementation relies on between organizations. The (manual) methodology proposed in 'operational definitions' of the domain-specific concepts. For [2] (and used as the methodological background in our research the brainstorming automation, these definitions, and also the project) is a guide for the integration and interpretation of the rules for their extension and interpretation, are supposed to TQM (Total Quality Management) instruments, aiming at BPR. compose the domain ontology and must be dynamically TQM is a team-based technology with techniques for creating created by the BPR team. effective teams, for organizing ideas (brainstorming, multivoting, 2. BPR ontology describes the BPR specific terms and affinity diagram, etc), for statistical analyses upon the target methodology (e.g. concepts for TQM statistical diagrams processes and upon the data collected during the BPR process. description and interpretation, for the diagram creation/ For the automation of this metodology, a virtual team of software modification, for the implementation of the methodology agents is intended, where the agents work either on behalf of the steps, etc). BPR ontology is predefined and should be users (BPR personal assistants and the mediator agent) or of other logically correlated with the domain ontology (e.g. the software agents (ontology agent) (see Fig. 1). At present, results correlation between certain results in the statistical analyses with respect to the application of multi-agent systems (MASs) to and the steps in the BPR methodology). BPR are not known in research and production. But, there are two 3. Communication ontology helps for the description and related domains where MAS applications are in progress: the interpretation of the communicative acts and of the dialog teamwork and workflow (i.e. organizational) technologies. between agents. The ontologies are a communication tool in a MAS, besides the All these ontologies are supposed to be explicit: declaratively transport protocol and the agent communication language. represented in an ontological knowledge base and managed by a Unfortunately, the ontology specification for MASs has not yet a dedicated ontology agent. standardized solution. So far, [3] is the most important specification that deals with ontologies for MASs. The automatic reasoning for BPR will mainly be with and upon the concepts in the three ontologies. Each user will be assisted in Section 2 reveals the types and roles of the ontologies for BPR the creation and interpretation of his ideas in terms of these automation. Section 3 enumerates the main requirements for ideas ontologies. mediation and for ontology integration. They refer to a BPR- dedicated MAS, but they could be extended to any teamwork- oriented one. Section 4 points out the coordinates of the upper- level ontology with linguistic features intended for ontology integration aiming at BPR. Personal Domain, BPR and Personal assistant communication ontologies assistant Ontology agent Mediator/ Facilitator agent Fig. 1 Types of agents and ontologies in a multi-agent system supporting a BPR team 3. MEDIATION OF IDEAS AND 4. TOWARD AN UPPER-LEVEL ONTOLOGY INTEGRATION FOR BPR LINGUISTIC ONTOLOGY During a brainstorming meeting, the ideas are required by the The practical reason for choosing natural language (NL) as human (and implicitly, the agent) mediator, with respect to the inspiration source for an upper-level ontology is its universality, topic of the meeting. Each member of the BPR team submits (by as well as its morphological and syntactic stability and, messages) his own ideas regarding that topic. The mediator agent implicitly, its integration ability. A linguistic ontology can be is in charge with the acquisition of the ideas and with their used for representing objects, processes and communicative acts, mediation and negotiation, in order to help the human mediator to as well as the correlation between them, in simple, compound and organize and synthesize them and to make appropriate decisions complex sentences, with the semantics and interpretation upon them. The reasoning that should support the mediation of borrowed from NL. The existing lexical ontologies, e.g. WordNet ideas must be directed to the ideas’ automatic comparison, and FrameNet, mainly emphasize the relations inside the lexical sorting, correlation, grouping, combination, negotiation. categories, without any concern about the composition, The reasoning for idea mediation will be simplified if the ideas interpretation and correlation of the sentences (i.e. of the ideas). are expressed in a predefined structured form (e.g. predefined Instead, the construction of the ontological sentences in the questionnaires) in terms of the vocabularies of the domain and intended upper-level ontology will follow five steps: BPR ontologies. However, this approach substantially restricts the creativity and innovation that are expected from the 1. Association of the words in the ontology’s vocabulary with brainstorming meetings. the morphological categories they belong to. In NL, the main morphological categories are nouns, verbs, adjectives, A much more free expression of the ideas in BPR entails the adverbs. Their counterparts in the intended ontology are: following requirements regarding the ontology representation and integration: • active objects, standing for the nouns directly involved in the verb’s action. • a BPR and, if possible, a domain ontology for both object • application specific or generic activities (operations), and processes, that is either (1) the extension of the existing standing for verbs. A particular kind of activities are the ontologies with explicit definitions and rules on activities communicative acts in the communication ontology. and processes (e.g. [6]) or (2) the combination of an object- • attributive objects, standing for adjectives (or other oriented ontology with a process-oriented one (e.g. KIF and linguistic categories with the role of noun modifiers); PSL (Process Specification Language) that have a similar • adverbial objects, standing for adverbs (or other underlying grammar); linguistic categories with role of verb modifiers). • the conceptual integration of the domain, BPR and communication ontologies; 2. Composition of the ontological simple sentences, by means of the syntactic roles of the objects relative to the action of • a representation language common to all three ontologies, the operations. At least three roles must be, implicitly or that is to be used as both content and communication explicitly, considered: agent (who produces the action), language for the agent implementation (e.g. the Interagent patient (object upon which the action operates) and recipient Communication Language in [4] that unifies the content and (receiver of the action’s result). the communication language). 3. Correlation of the elements inside or between the The ontology integration could be achieved either morphological categories by semantic relations (e.g. • by an upper-level ontology; or synonymy, antonymy, hyponymy, hypernymy, meronymy, • by a translation algorithm between ontologies [3]. holonymy etc, see WordNet [5]). The disadvantage of the latter is that this algorithm will be mostly 4. Correlation of the activities (i.e. verbs) by intersentential encoded and the ontology integration cannot be performed in the relations across simple sentences. These relations help us conception phase, as recommended. build the ontological compound or complex sentences. 5. Correlation of the objects in different sentences by coreferences (anaphoric references). NL simple sentence, used as inspiration source for building the ontological simple sentence, is summarized below: Simple sentence = Subject + Predicate, with (I) inference upon them can be consistently transposed to first order Subject ∈{Noun Phrase, Noun Substitute, Verbal Phrase, predicate calculus [1]. Like in NL, the extension of the ontology Clause, Indefinite/ Formal Subject} is natural, by new types of objects, operations, roles and relations. The ontological sentences will also be used for assisting the users Predicate = Finite-Verb + Verb Determiners + Adverbial in expressing their ideas on objects and processes. The syntactic Modifier(s) and semantic roles and relations facilitate the sentence where Verb Determiners are: comparison, classification and integration, i.e the integration of Link-verb + Subject Complement / the ideas they represent. Direct Object / Direct Object + Direct Object / Indirect Object + Direct Object / 5. CONCLUSIONS Prepositional Object / This position statement focused on the need for a new type of Object + Object Complement mediation in a MAS, namely the mediation of ideas (instead of Subject/ Object Complement gives information on the subject/ services as provided by the existing facilitators: matchmakers or object. Link-verb is a finite verb that expresses being/ passing/ brokers). It has arisen during the requirements analysis for the remaining/ seeming or appearing in a certain state. Indirect/ automation of a modern BPR methodology, where the Direct Objects are in dative/ accusative case. brainstorming sessions have a central place. The semantic networks combined with the semantic roles in the Ideas mediation and the ontology integration requirements entail case grammars have been proved to help for the translation of NL the need for an upper-level ontology with linguistic features, as morphology, syntax and semantics into a stylized form of NL, this statement tries to motivate. The main steps in the construction without the ambiguities of the pure NL [1]. of a linguistic ontology and its general advantages are briefly presented at the end of the statement. The intended ontological simple sentence, that abstracts the pattern (I) above, will have the format: Regarding the implementation of the ontologies in a MAS for BPR, the conclusion is that the standardization and integration of Ontological Simple Sentence = Agent Phrase + Operation Phrase the two technologies (software agents and formal ontologies) are Agent Phrase = AGNT + Object Phrase still in incipient phases and do not encourage and help their use in Operation Phrase = (OPERATION) + BPR automation. + { + Active Object Phrase} ... +{ + Adverbial Object Phrase} Object Phrase = [Object_Type: Individual]+ REFERENCES + Object Determiner(s)+ Object Modifier(s) [1] Allen J., Natural Language Understanding. Benjamin/ The ontological simple sentence will be used to define and describe BPR and domain specific objects and activities, generic Cummings Publ. Comp., USA, 1987, 1995 operators (e.g. for object and activity definition, qualification, [2] DON, Handbook for Basic Process Improvement. semantic correlation) as well as the communicative acts. So far, Department Of The Navy (DON), USA, 1996 the intended communicative acts are 'query', 'reply' and 'notify' (or 'inform'). [3] FIPA, Ontology Service Specification. Geneva, 2000 In NL, the compound sentence joins independent simple [4] Martin D., Cheyer A., Moran D., The Open Agent sentences by coordinating conjunctions (copulative, disjunctive, Architecture: A Framework for Building Distributed adversative, resultative, explanatory conjunctions) or adverbs or Software Systems. Intl. Journal "Applied Artificial asyndetically (without conjunctions). The complex sentences are Intelligence", vol. 13, No. 1-2, 1999 composed of dependent (subordinated) sentences (noun/ adverbial/ relative/ appositive subclauses) correlated to a main [5] Miller G., WordNet: A Lexical Database for English. sentence (clause). Communication of ACM 38:11, 1995 Examples of intersentential relations for constructing ontological [6] Uschold M., King M., Moralee S., Zorgios Y., The compound sentences (practically, relations between activities Enterprise Ontology. Knowledge Engineering Review, defined and described by simple sentences) are: ‘and’, ‘or’, 'not', vol. 13, 1998 ‘cause/ for’, etc. In a complex sentence, the activities must be correlated by subordinating relations like: ‘if-then-else’, 'while', subordinating 'cause', 'event' ‘purpose’, ‘consequence’, ’before’, ‘after’ , 'case', etc. For each kind of user (member or mediator) and for each BPR step, the BPR ontology comprises a scenario like an ontological complex/ compound sentence. The predefined ontological (simple and compound/ complex) sentences represent the set of axioms upon the objects/ processes/ communicative acts in the three ontologies. Their logic and the