Introduction to Conceptual Modeling Gabriela P. Henning INTEC (Universidad Nacional del Litoral - CONICET) 3000 - Santa Fe, Argentina 1 Introduction to Conceptual Modeling - Outline Motivating questions Knowledge representation and reasoning Critical issues Knowledge engineering Emerging paradigms in the 70 and 80 Current trends in knowledge representation: Conceptual modeling today 2
Questions What is a model? Are there different types of models? What is conceptual modeling? Why do we need explicit models in Computer Science? Which are the differences among data, information and knowledge? Different types of information/knowledge knowledge? Extensional vs. intensional information Declarative vs. procedural knowledge Particular vs. general information 3 Questions What is a model? Are there different types of models? What is conceptual modeling? Why do we need explicit models in Computer Science? Which are the differences among data, information and knowledge? Different types of information/knowledge knowledge? Extensional vs. intensional information Declarative vs. procedural knowledge Particular vs. general information 4
Introduction to models Human beings have used symbols and representations to model their environment since the beginning of civilization 5 Models A model is always an abstraction of reality Model is a widely used term The term model can be interpreted in different ways by distinct communities There are models of physical things (models of entities and systems having actual, real existence) and models of insubstancial (man-made) systems, such as: Conceptual models Causal models Data models Statistical models Business process models Architectural models.. 6
Questions What is a model? Are there different types of models? What is conceptual modeling? Why do we need explicit models in Computer Science? Which are the differences among data, information and knowledge? Different types of information/knowledge knowledge? Extensional vs. intensional information Declarative vs. procedural knowledge Particular vs. general information 7 Conceptual Modeling According to John Mylopoulos (1992) the discipline of conceptual modeling is: the activity of formally describing some aspects of the physical and social world around us for purposes of understanding and communication. Conceptual modeling supports structuring and inferential facilities that are phychological grounded. After all, the descriptions that arise from conceptual modelling activities are intended to be used by humans, not machines The adequacy of a conceptual modelling notation rests on its contribution to the construction of models of reality that promote a common understanding of that reality among their human users. 8
Conceptual Modeling The specification of a conceptual model can be viewed as a description of a given subject domain. This is why conceptual models are also known as domain models. The aim of a conceptual model is to explicitly express the meaning of terms and concepts used by domain experts to discuss the problem, and to find the correct relationships between different concepts. A conceptual model attempts to clarify the meaning of various, usually ambiguous terms, and ensure that problems with different interpretations of these terms and concepts cannot occur. 9 Conceptual Model A conceptual model must be explicitly chosen to be independent of design, implementation concerns (e.g., concurrency issues) or technological choices (e.g. data storage technology), that should influence the particular applications or telematic systems based on such model. Conceptual specifications are to be used to support understanding (learning), problem-solving, and communication among stakeholders about a given subject domain. Once a sufficient level of understanding and agreement about a domain is reached, then the conceptual specifica_ tion becomes a basis for subsequent development of applications in the domain. 10
Conceptual Modeling Ullmann s triangle: the relations between a thing in reality, its conceptualization and a symbolic representation of this conceptualization. Concept (Conceptualization) Note de dotted line between represents language and reality. It indicates that the relation between them is always established by the intermediation of a certain conceptualization Symbol (language) refers to abstracts Thing (reality) 11 Distinction between a model and its interpretation Conceptualization representedby interpretedas Modeling Language instanceof usedto Compose instanceof usedto Compose Model representedby interpretedas Model Specification Guizzardi, 2005 12
Distinction between a model and its interpretation A conceptualization is the set of concepts used to articulate abstractions of state of affairs in a given domain. The abstraction of a portion of reality articulated according to a domain conceptualization is termed here a model. The representation of a model in terms of a language is called a model specification, or simply specification. The language used for the creation of a specification is called a modeling language. 13 Distinction between a model and its interpretation A language can be seen as determining all possible specifications (i.e. all grammatically valid specifications) that can be constructed using that language. A conceptualization can be seen as determining all possible models (standing for the state of affairs) which are admissible in such domain. Guizzardi defends the precedence of real-word concepts over formal ones and implementational issues in the design/adoption of conceptual modeling languages. He points out the importance of the so-called domain appropriateness and comprehensibility appropriate_ ness of languages. 14
Distinction between a model and its interpretation The domain appropriateness of a language is a measure of its suitability to model the phenomena in a given domain. In other words, it can be seen as the truthfulness of the language to a given domain or reality. The comprehensibility appropriateness of a language refers to how easy if for a user of the language to recognize what that language s constructs mean in terms of domain concepts. Moreover, it refers to how easy is to understand, communicate and reason with the specifications produced in such language. Both domain appropriateness and comprehensibility appropriateness are properties of the represents relationship in Ulmann s triangle. 15 Questions What is a model? Are there different types of models? What is conceptual modeling? Why do we need explicit models in Computer Science? Which are the differences among data, information and knowledge? Different types of information/knowledge knowledge? Extensional vs. intensional information Declarative vs. procedural knowledge Particular vs. general information 16
Early Scientists Thoughts.. 17 Why do we need explicit models?.. Need to explicitly represent knowledge 18
Why do we need explicit models?.. To build systems exhibiting some kind of intelligent behavior. Many of the problems that computers are expected to solve require extensive and explicit knowledge about the world of study: objects, properties, categories and relations between objects; situations, events, states and time; causes and effects, etc. To capture the relevant aspects of some world, so the model can serve as a point of agreement among members of a group, and to communicate that common view to newcomers. Because explicit models are useful in rationalizing and supporting information system development. To represent requirements to be considered during the early phases of system development. As a foundation for the integration of different system applications. 19 Questions What is a model? Are there different types of models? What is conceptual modeling? Why do we need explicit models in Computer Science? Which are the differences among data, information and knowledge? Different types of information/knowledge knowledge? Extensional vs. intensional information Declarative vs. procedural knowledge Particular vs. general information 20
Data, information, knowledge.. The Information Pyramid / The Knowledge Hierarchy Decision Making Collecting Filtering Synthesizing Analyzing Summarizing Organizing Wisdom/ Intelligence Knowledge Information Data Signals 21 Data, information, knowledge.. Data Raw signals in digital form. Many times obtained by processing signals from sensors, bar code readers, etc. A collection of symbols without any meaning beyond its existence. PO30478 500C Information A set of data which have been given a meaning by formulating relations between the data elements in a given context. Meaning attached to data Understandable by humans and computers S O S PO30478, P1159, 20, 30-04566291-3, 03/11/10,.. 22
Data, information, knowledge.. Knowledge Constitutes a collection of information with the intention of a certain kind of use, May attach purpose and competence to information New knowledge may be created from existing knowledge by using inference processes. Has potential to generate action If (Reactor.temperature Reactor.setpoint) > 10 Then Reactor.status = RunawayAlert If Reactor.status = RunawayAlert start ShutdownProcedure Understanding or reasoning refers to an analytic and cognitive process, which takes some knowledge as its input to infer new knowledge as its output by some kind of interpolation 23 Data, information, knowledge.. Intelligence Intelligent systems Computational systems that are capable to solve problems or do things that require intelligence when done by humans. Many of nowadays intelligent systems use an explicitly represented store of knowledge to reason by considering goals, the environment, other computational agents, etc. There are many particular traits, behaviors or capabilities that researchers would like an intelligent system to display, such as: deduction, induction, reasoning, problem solving, planning, learning, knowledge representation, natural language processing, motion and manipulation, perception, social intelligence, etc.. 24
Questions What is a model? Are there different types of models? What is conceptual modeling? Why do we need explicit models in Computer Science? Which are the differences among data, information and knowledge? Different types of information/knowledge knowledge? Extensional vs. intentional information Declarative vs. procedural knowledge Particular vs. general information 25 Extensional vs. Intentional Information An extensional definition of information would define things or concepts by listing everything that falls under such definition. Examples: An extensional definition of mother would be a listing of all women that are mothers in the world. Similarly, the extensional definition of bachelor would be a listing of all the unmarried men in the world. An intentional definition of information would define the meaning of a term by specifying all the properties required to come to such definition, that is, the necessary and sufficient conditions for belonging to the set being defined. Examples: An intensional definition of mother is woman with one or more children. An intentional definition of "bachelor" is "unmarried man." Unmarried man is a necessary and sufficient property that defines a bachelor. 26
Extensional vs. Intentional Information To distinguish between extension and intension, let s analyze a predicate, like the English word red. Two meanings can be given to it: a) The set of all red things this is called the extension of the predicate b) An abstract entity which in some sense characterizes what it means to be red. It refers to the notion of redness, which may or may not be true of a given object this is called the intention of the predicate. In many philosophical theories the intention of a predicate is identified with an abstract function which applies to possible worlds and assigns to any such world a set of extensional objects, i.e. the intention of red would assign to each possible world a set of red things. 27 Declarative vs. Procedural Knowledge Declarative representations have knowledge in a format that may be manipulated, decomposed and analyzed by various reasoning tools (i.e., reasoners). Declarative representations are associated with know that or know what. Clear advantages of a declarative representation are: a) the ability to use knowledge in ways that the system designer did not foresee, and b) the possibility of reusing the representation for different purposes. Procedural representations encode knowledge in a way that is linked to how to achieve a particular result. Procedural knowledge, also known as imperative knowledge, is the knowledge put into effect in the execution of some task. Procedural representations are associated with know how. 28
Particular/Specific vs. General knowledge Specific knowledge can be regarded as knowledge that is costly to be transferred among different agents. It can be seen as case-specific or situation dependent knowledge. General knowledge can be regarded as knowledge that is inexpensive to transmit due to its generality. It can be seen as knowledge that transcends or goes beyond specific situations. It is always desirable to extract general knowledge out of specific one. One possible mechanism can be inductive generalization. It proceeds from a premise about a sample to a conclusion about the whole population. 29 Introduction to Conceptual Modeling - Outline Motivating questions Knowledge representation and reasoning Critical issues Knowledge engineering Emerging paradigms in the 70 and 80 Current trends in knowledge representation: Conceptual modeling today 30
Knowledge Representation & Reasoning Knowledge Description of the world of interest that is usable by machines to draw conclusions about such world The psychological result of cognitive processes, i.e., of perception, learning and reasoning. That which is understood or can be understood The wing wherewith we fly to heaven (Shakespeare) Knowledge differs from data or information in that new knowledge may be created from existing knowledge using inference processes. Reasoning Way of thinking that is coherent and logical Logical inference process The process of creating new knowledge from existing one 31 Knowledge Representation & Reasoning Knowledge representation and reasoning is an area of artificial intelligence whose main goal is to represent knowledge in a manner that facilitates inferencing (i.e. drawing conclusions) from knowledge. It analyzes how to formally think - how to use a symbol system to represent a domain of discourse, along with functions that allow inference. Representation of knowledge Description of the world of interest that is usable by machines to draw conclusions about such world Reasoning based on explicitly represented knowledge Working hypothesis: Knowledge of the world can always be articulated and used as needed. 32
Some knowledge representation issues What form is the knowledge to be expressed? How can a reasoning mechanism generate new knowledge? How can knowledge be used to influence a system s behavior? How is incomplete, inconsistent or noisy information properly handled? How can practical results be obtained when reasoning is intractable due to the complexity of the domain?.. 33 KR&R Knowledge Representation How information can be appropriately encoded and utilized in computational models of cognition? Two primary areas of activity: Designing formats for expressing information Mostly "general purpose" representation languages (e.g., first order logic) Encoding knowledge (knowledge( engineering) Mostly identifying and describing conceptual vocabularies (ontologies) Declarative representations are the focus of KR technology Explicit knowledge that is domain-specific but task- independent. Separating that/what from how. 34
KR&R Reasoning Computational methods for creating new knowledge and information from existing knowledge Very general methods: e.g. modus ponens from first order logic Task-specific specific methods: algorithms for planning, scheduling, diagnosis, constraint satisfaction, etc. Methods for managing reasoning: e.g., hybrid reasoning, parallel processing, etc. Analysis of the reasoning capabilities Examination of properties such as soundness, completeness, complexity, etc. Methods for creating explanations from the obtained reasoning results, e.g. explanation of the line of reasoning. 35 Knowledge representation & reasoning Expressiveness vs. tractability (effective reasoning) trade- off How to express what we know? How to reason with what we express? Every representation ignores something about the world When modeling the real world, KRs are always imperfect, i.e. KRs are surrogates for the real world Given a KR, there are two questions to ask: Semantics -- For what is it a surrogate? Fidelity -- How accurate is it? 36
Introduction to Conceptual Modeling - Outline Motivating questions Knowledge representation and reasoning Critical issues Knowledge engineering Emerging paradigms in the 70 and 80 Current trends in knowledge representation: Conceptual modeling today 37 Knowledge Engineering KE Early definition: KE is an engineering discipline that involves integrating knowledge into computer systems in order to solve complex problems, normally requiring a high level of human expertise (Feigenbaum & McCorduck, 1983). Nowadays, KE refers to the building, maintaining and development of knowledge-based systems that can be used in many computer science domains, such as artificial intelligence, database development, data mining, intelligent systems, decision support systems and geographic information systems, among others. 38
Knowledge Engineering Can be defined as the process of defining the scope of a knowledge-based system, eliciting, capturing, structuring, formalizing, validating and verifying, operationalizing information and knowledge involved in a knowledge-intensive problem domain, in order to construct a program/system that can perform a difficult task/set of tasks adequately. 39 Knowledge Engineering Can be defined as the process of defining the scope of a knowledge-based system, eliciting, capturing, This is not a sharp list. These phases generally overlap, the structuring, whole process might be iterative, formalizing, and many challenges could appear validating and verifying, operationalizing information and knowledge involved in a knowledge-intensive problem domain, in order to construct a program/system that can perform a difficult task/set of tasks adequately. 40
Problems in Knowledge Engineering Complex information and knowledge are difficult to observe/elicit, make explicit, comprehend and capture Experts and other sources generally differ on their views Multiple knowledge sources which coexists have intrinsic different information structures : textbooks graphical representations heuristics Skills Knowledge is valuable and often outlives a particular implementation. Knowledge is not static Need for knowledge management and maintenance tools Errors in a knowledge-base can cause serious problems 41 Issues in knowledge engineering There are: Different types of knowledge, and that influences the right approach and technique that should be used for the type of knowledge being required. Distinct ways of representing knowledge (different languages and formalisms), which can aid the acquisition, validation, and re-use of knowledge Different types of experts and expertise, such that methods should be chosen appropriately. Distinct goals drive the development of intelligent/expert systems. 42
Introduction to Conceptual Modeling - Outline Motivating questions Knowledge representation and reasoning Critical issues Knowledge engineering Emerging paradigms in the 70 and 80 Current trends in knowledge representation: Conceptual modeling today 43 A Short History of Knowledge Systems general-purpose search engines (GPS) first-generation rule-based systems (MYCIN, XCON) emergence of structured methods (early KADS) mature KE methodologies (CommonKADS) Ontologies 1965 1975 1985 1995 2000-2010 => from art to somehow discipline => 44
A Short History of Knowledge-based Systems 45 Early history of knowledge representation: 60 & 70 Origins: Problem solving work at MIT, CMU, (Stanford) Driven by natural language understanding Many Ad-hoc formalisms Procedural vs. Declarative knowledge controversy Informal semantics Problems: How do we assign meaning to things? How/when can we say that a computer understands? 46
Emerging paradigms in the 70 & 80 Predicate logics Semantic nets Frames Production rules 47 Emerging paradigms in the 70 & 80 Predicate logics Semantic nets Unstructured node-link graphs No semantics (minimum) to support interpretation No axioms to support reasoning capabilities Frames Production rules 48
Semantic nets Semantic memory motivation Quillian, 1966 Understand the structure of human memory, and its use in language understanding What sort of representational format can permit the meanings of words to be stored, so that humanlike use of these meanings is possible? Psychological evidence that memory uses associative links in understanding words. Claim that people use same memory structure for a variety of tasks 49 Semantic nets Directed, labeled graphs used to represent concepts and the relationships between them. Arcs define binary relationships that hold between the objects that define the nodes. 50
Semantic nets The ISA and AKO relationships were sometimes used to: Link a class and its superclass Link a class with its instances Some links are inherited along ISA paths (e.g. has part relationship ) The semantics can range from very formal (Krypton), to formal (KL- ONE), and informal. It depends on the implementation. 51 Semantic nets - Reification Non-binary relationships can be represented by turning the relationship into an object : Logicians call this issue reification Reify v: Consider the abstract object v to be real Recipient Give Object Object Peter Giver Giver Chemistry book Hans 52
Semantic nets Classes and instances Many semantic nets distinguish: Nodes representing classes and instances The subclass relation from the instance-of link 53 Emerging paradigms in the 70 & 80 Predicate logics Semantic nets Frames Structured semantic nets Object-oriented description Prototypes Class-subclass subclass taxonomies Production rules 54
Motivations for frame-based representations Minsky s original motivations and observations: Famous analysis of a birthday party. An attempt to model of human cognition (the structure of knowledge memory) and some foundations for common sense reasoning (e.g. the capability to represent things like a room, an animal, etc.). Memory is full of prototypical situations, richly interconnected. A frame-based representation is organized around prototypes. Semantic networks evolved into frames. Frames have a less shallow structure than semantic networks. A frame may contain information about the components of the concept being described, links to other concepts, as well as procedural information on how the frame can be accessed and change over time. 55 Frames A frame is similar to the notion of object in OOP, but has more metadata, and a primitive notion of behavior. A frame has a set of slots or properties A slot represents a relation to another frame (or to a value) A slot has one or more facets A facet represents some aspect of the relation Some facets have procedural capabilities, behaving as demons In some systems, the slots themselves are instances of frames. In others, slots may contain methods. Frame systems support inheritance.. Issue: Simple vs. Multiple inheritance 56
Frames A slot in a frame holds more than a value or a set of values. Facets participate in the specification of slots. Facets may include: Current fillers (e.g., values) Default fillers Cardinality: minimum and/or maximum number of fillers type restriction on fillers (valuetype or valueclass: usually expressed as another frame object) constraints on the inheritance mechanisms (inheritance roles) Demons (attached procedures) that are triggered when something changes in the slot values (if-added, if-removed, etc.) Salience measure (for inference mechanisms) Attached constraints or axioms 57 Frames 58
Frames 59 Frames 60
From frames to description logic There is a family of frame-like representation systems with a formal semantics: : e.g. KL-ONE, LOOM, et. An additional thing that can be done with these systems is automatic classification: Finding the right place in a hierarchy of objects (taxonomy) for a new description. There is a need to keep the language simple so as to ensure that all inferences can be done in polynomial time. Ensuring tractability of inference 61 Emerging paradigms in the 70 & 80 Predicate logics Semantic nets Frames Production rules Situation-action rules: IF (warning-light on) THEN (turn-off unit) If-then inference rules: IF (warning-light on) THEN (reactor overheating), IF (warning-light on) THEN (reactor overheating) 0.95) Hybrid procedural-declarative representation Basis for first generation of expert systems 62
Introduction to Conceptual Modeling - Outline Motivating questions Knowledge representation and reasoning Critical issues Knowledge engineering Emerging paradigms in the 70 and 80 Current trends in knowledge representation: Conceptual modeling today 63 Knowledge representation in the 00 s Web-based based systems Driven by new classes of applications (e.g. e-e commerce, information retrieval on the Web, Web services, etc.) Incorporation into traditional applications Support to Software Engineering, collaborative design process, requirements engineering Support for information integration processes Business process representation Support of business process reengineering Ontologies!! 64
References Brachman, R. J. The future of knowledge representation, Proceedings of AAAI-90, 1084 1092, 1990. Davis, R.; Shrobe H.; Szolovitz P. What is a knowledge representation. AI Magazine 14, 17 33, 1993. Guizzardi, G. Ontological Foundations for Structural Conceptual Models. CTIT PhD Thesis Series, No. 05-74, Universiteit Twente, Enschede, The Netherlands, 2005 Minsky, M. A Framework for representing knowledge. In: Brachman, R.J.; Levesque, H. (Eds.) Readings in Knowledge Representation. Morgan Kaufmann, San Mateo, California, 1985. Mylopoulos, J. Conceptual Modeling and Telos. In: Loucopoulos, P. and Zicari, R. (Eds), Conceptual Modeling, Databases and CASE, Chapter 2, 49-68, Wiley, 1992. Mylopoulos, J. Conceptual Modeling Information Modeling in the Time of the Revolution, Information Systems 23 (3-4), June 1998. Olivé, A. Conceptual Modeling of Information Systems, 2007. Woods, W. A. What s in a link: Foundations for semantic networks. In: Bobrow, D.G., Collins A. M. (Eds.), Representation and Understanding: Studies in Cognitive Science, Academic Press, New York, 35-82, 1985. 65