METADATA STRUCTURES and PROGRAMMING for DISTRIBUTED DICTIONARY RESOURCES in a CONTEXT of LEARNING Volodumur Shirokov, Volodumur Manako Ukrainian Lingua-Informational Fund of the National Academy of Sciences of Ukraine 54, Volodimirska str., Kiev, 01601, Ukraine, tel. (380)(44) 267-4859, e-mail: vlad@umif.kiev.ua Katherine Sinitsa, Alla Manako International Research and Training Center of Information Technologies and Systems UNESCO/IIP, 40, prosp. Glushkova, 03187, Kiev, Ukraine, tel. (380)(44) 266-6311, e-mail: kathy@tel.dlab.kiev.ua, alla@tel.dlab.kiev.ua Abstract. A model to support Autonomous Learning by learning-centric Dictionaries (ALD model)is suggested. Dictionary metadata structures and programming based on ALD model as well as a unified dictionary player and some case studies are described. Анотація. У доповіді наведена наша точка зору на роль словникових ресурсів у розподілених електронних/цифрових середовищах у контексті навчання і структури метаданих та програмування для цих ресурсів. 1.0 Introduction Present trends in the job market require people to adapt their knowledge and skills to changes in the workplace, thus, making it crucial to have access to life long autonomous learning (EC’95g). The old metaphor restricted the role of a dictionary to the auxiliary storage of specific portions of information that can be retrieved upon a direct user's request. The role of interactive dictionaries may be extended to the one of a tool which supports various learning activities, facilitating the formation of learner’s mental knowledge structures, suggesting a conceptual framework for domain knowledge, ensuring the incremental incorporation of new knowledge guided by the learner's professional goals (Colazzo, 1996), (Sinitsa&Mizoguchi, 1998), (ComNEd'99), (EDT, 1999). A forefather of Russian academic science, M. Lomonosov recommended to learn mathematics, because “it puts one’s mind into order”, so could good interactive dictionaries. Design and maintenance of the dictionary resources require extensive programming skills and enormous manpower to create and manage a variety of software tools for the identification, extraction, ordering, transfer, storage, maintenance, representation of dictionary resources. Much of the progress of modern software development is based on the concepts of “interoperability”, “adaptability” and “reusability”. Interoperability of dictionary resources is defined as an ability to ensure a usage of dictionary resources components developed in some place by some set of tools or for some platform, in another place, with a different set of tools or at other platforms. Adaptabity of dictionary resources is defined as an opportunity to tailor dictionary resources content to individual and situational needs. Reusability of dictionary resources is defined as a feature that allows for incorporation of dictionary resources components into multiple applications. Interoperability and reusability requires technical, economic, and legal infrastructures. Of these three conditions, the technical infrastructure is almost certainly the easiest -- agent-based systems and computing, functional specifications, component architecture, metadata, object-oriented programming, and true mark-up languages are likely to play an important role. However, there are not enough “common solutions” applicable to the process of transformation of the dictionary resources, dictionary industry and infrastructure from a paper-based to a “digital-based” environment (WFC, 1996), (CSM, 2000) to ensure a discovery and use of interoperable, adaptable and reusable dictionary resources in distributed digital environments. In this paper a model to support Autonomous Learning by learning-centric Dictionaries (ALD model) is described together with dictionary metadata structures and programming based on ALD model. It also contains a description of a unified dictionary player and some case studies. 2.0 Background Approaches for discovery/use of such“common solutions” may be roughly represented as a decompositioncomposition approach and “learning-centric” approach. Decomposition-composition unformal approach. The POINTER EC Project ( POINTER) defines terminological resources as: “ a structured set of concepts and their designations (graphical symbols, terms, phraseological units, etc.) in a specific subject field”. For the created terminological resources and their components some solutions are formulated unformally. Decomposition-composition formal approach. Dictionary resources are created by decompositions of general formal model into separate solutions with later compositions of those solutions for a various domains (MOV, 1996), (ModSUM, 1996), (DICsit, 1998). Common formal structural model of lexicographical (including, dictionary) systems is constructed on the basis of the principles formulated for Language science and modularization of dictionary metadata (ShirCrimea, 1994), (ShirMON, 1998), (UkrPROG98), (ShirMMS, 1999), (Crimea 99), (CSM, 2000). Learning-centric approach. Taking into account a general model of the lexicographic systems, dictionary resources are created and used in a distributed digital environment in a broader context - a context of learning (ComNEd'99), (HCI, 1999), (Crimea, 1999), (CSM, 2000). Some key components of the dictionary environments are described in (UkrINTEI, 1999), (Crimea, 1999), (CSM, 2000). 3.0 Developing a Learning-Centric Approach 3.1 A common model to support autonomous learning by learning-centric dictionaries We consider a user of the dictionary resources as an "autonomous learner" - an Agent (person, family, etc.) that self-directed discover and use resources in electronic environments, trying to enlarge his/her knowledge, skills or capabilities for professional and other activity. Thus, the autonomous learner must (ComNEd'99), (HCI, 1999): (1) discover/use personal learning objects: learning strategies, goals, action plans taking into account both personal (learning style, skills cards, resources, etc.) and broader (legal, social, economic, technical) context; (2) discover/learn from personal or external resources. Learning Virtual Environment (LVE) organized as a set of interoperable and communicating networked tools/systems, may add value to each of its components by extension and integration of their Roles in autonomous learning process, introduction of cognitive support for a autonomous learner. Taking into account (1)-(2), a model to support autonomous learning by LVE can be defined as : <TEXT> Annotation: Annotation Name for <TEXT>: <al> /* (i.e. autonomous learning) Short Annotation Description of <al>: model to support autonomous learning by LVE. Description Level: <level: al> Full Description of <al>: al.1: al.2/2: al.3: for <Right> <personal/broader context> <to support>/<do> the <Right> <autonomous learning> al.4: al.5: al.6: al.7: of the <<Right> <learning resources> at the <Right> <time>, at the <Right> <place>, at the <Right> <cost> Some of rules to construct such <TEXT> are: the Roles of autonomous learner are supported by a Role-playing Agents (UkrINTEI, 1999) according to this <TEXT> (Examples of Roles are shown in the Table 1). Autonomous learner is a Roles-playing Agent; Root(s) for LVE/( Autonomous Learner) Role-play of <level: rl> are: rl.*; <TEXT> is not determined ab inito the Action order for <al>; <TEXT> interpretation in digital environment: <...> is container abstractions for digital objects/actions metadata with complex distributed active relationships (WFC, 1996), (DARs, 1997), (CSM, 2000); each <Right> (or <Best> for an Agent point of view) has various behaviors in autonomous learning lifecycles (see detail about behavior “states” of dictionary resources in (UkrINTEI, 1999)). It could be measured in personal/broader dimensions and scales. For example, <Right> for <learning resources> is an Agent, which supports the meeting of the following conditions: (1) <learning resources> must correspond to standards; (2) <learning resources> must have definite multilingual extensions, etc. Decomposition of this <TEXT> is performed as follows. We define <<Right> <autonomous learning> as <<Right> <sequences> of <autonomous learning activity>>, which an autonomous learner discovers, explicitly defines, orders, performs, etc. In a “classroom” LVE, keeping balance between new and known components, defining granularity of learning units, planning, sequencing and monitoring is done by a tutor, thus allowing a learner to concentrate on subject matter. An autonomous learner has to perform both roles simultaneously, which requires reasoning at both subject level and meta-level. Examples of Core activities in autonomous learning are shown in the Table 1 (see also details in (ComNEd'99), (EDT, 1999), (UkrINTEI, 1999)), Activity Roles Details Comment + creator + digital traveller + learner of a subject + evaluator + manager ... Personalization of learning + strategies/ goals/ plans + skills, capabilities + learning style + learning resources ... Navigation + in knowledge space + discovery/use related learning blocks Manage-ment + educational of learning + informational + administrative ... of learning strategies, goals, action plans perform navigation in knowledge space mastery of learning objects, of performance content (history, current, new) of selflearning Action Example of personalization: for <personal/broader context> <do> the <<Right> autonomous learning> <of highest quality of terminological resources of the Internet>, <tailored to my individual and situated needs>, <delivered costeffectively> <to my home/mobile workplace> and <on monthly basis> where to get relevant information, how to process it and adopt to new situations parameters for personal profiles: life long learning autonomous learning collaborative work subject learning ... Table 1 - Examples of Core activities in autonomous learning Therefore, in accordance with the Table 1, <<Right> <autonomous learning>> is defined as: <level: al.3>: <<Right> <sequences> of: <Right> < Role> <Right> <Personalization of self-learning > <Right> < Navigation > <Right> < Management of self-learning >>, etc. A note on usage of the term “digital traveller” in the Table 1: “Ricky Erway's suggested a Digital Tourist metaphor (OCLC). His metaphor casts the resource seeker as a "digital traveler" who, like a real traveler, can navigate in early stages of the process using a general "phrase book" type of approach. However, just as a real traveler must actually learn the language of the specific culture in order to navigate within it, the information seeker must adopt domain-specific syntax and semantics to succeed in the discovery process, i.e. change a role of "digital traveler" to the one of “a learner of a subject”). Further details of <Right> <learning resources >. Interactive dictionary is a collection of “right created” (POINTER) knowledge structures/ objects, which are relatively stable in many dimensions: time, space, used, etc . (Crimea, 1999). THUS: dictionary knowledge structures/objects may serve as life long learning resources for autonomous learner. For example, a learner discovers or studies some of terminological resource in the Internet. A knowledge object - the Term ‘Internet’ is presented in the dictionaries as TermDesignation & TermDescription. TermDesignation (POINTER) for term ’Internet’ is a stable designation - <Internet>. Therefore, if one stores TermDesignation together with some personal metadata (personal performance content metadata ) about this term in his/her Personal library, then this TermDesignation will be a long-term “Entry” for life long autonomous learning of this term; dictionary knowledge structures/objects (for example, the term ‘Internet’) may interact via Agents with the related knowledge structures of glossaries (containing this term), which are embedded in or linked to learning resources. Those knowledge structures/objects of glossaries may have links to the learning blocks (about ‘Internet’) of those resources. Thus, dictionary knowledge structures/objects may be used by a tool as a life long road-map to generate a plan for autonomous learning of the related learning resources. Taking into account those reasons, <<Right> <learning resources>> is defined as: <level: al.4>: <Right> <<<stable> <right created> <knowledge structures/objects collections of domains>> and <related learning resources >>. Taking into account the above, a model of autonomous learning support by dictionaries is defined as the following <TEXT> (In this paper a full <TEXT> is not included): <level: al.*> <for <personal/broader context> <to support>/<do> <Right> <sequences> of <Right> < Self-role(s)-play> <Right> <Personalization of self-learning> <Right> < Navigation >; <Right> < Management of self-learning> of the <Right> <stable> and <right created> <knowledge structures/objects collections of domains> and <related learning resources >, at the <Right> <time>, at the <Right> <place>, at the <Right> <cost>. Interactive dictionaries, which support this Model and are naturally added/embedded in LVE, is called here learning-centric dictionaries (LDS). The virtual environment, which supports core LDS functions is called a Dictionary Player (DP). Such full <TEXT> we call model to support autonomous learning by learning-centric dictionaries (ALD model). 3.2 Towards Learning-Centric Interactive Dictionaries Currently, all spectrum of dictionaries exists in an electronic form: from short glossaries, to multilingual dictionaries, explanatory dictionaries and encyclopedia. Dictionaries are created by context-dependent methods and procedures, they have complex knowledge structure, diverse ordering systems, context-dependent behavior, etc. Modern computer-based technologies freed these products from space and time restrictions, added new media and dynamic objects, connected knowledge objects by distributed and dynamic relationships. Interactive dictionaries vary in size, purpose, explanation style, interface, access to entries, and use of illustrative media, but have some specific features in common. These features, important for the dictionary support of autonomous learning, are the following. Interactive dictionaries are stable collections of formally defined and viewed from multiple points knowledge structures/objects for autonomous learning. The dictionary knowledge structures/objects are formally defined by dictionary mark-up language (ShirMON, 1998) and represented by information fragments collections (CSM, 2000). Some fragments have names indicating their content, which are used in navigation; some fragments collections may have their own catalogs; some fragments may have complex relationships and links to each other. Dictionary knowledge structure consists of a large number of sub-structures, which are multi-dimensional, colorful and allows for several views. Thus, a dictionary being a specific sublimation of knowledge, should be able to demonstrate its content from different viewpoints. Few tools, which support autonomous learning of dictionary knowledge. Interactive dictionaries do not support an explicit order for autonomous learning of dictionary knowledge, i.e. they do not identify appropriate fragments of dictionary knowledge, put pieces together and evaluate the learning solutions. Navigation process in dictionary knowledge structures/objects is long and complicated, and must be actively present in a “memory of the autonomous learner”. Moreover, autonomous learning of dictionary knowledge requires decomposition/composition of various iterations and/or cycles: goals - objectives - plans - performance evaluation (ComNEd'99), (EDT, 1999). For example, autonomous learning of dictionary knowledge requires (HCI, 1999), in particular a local learning (mastering a specific concept, extending and upgrading one's knowledge of a separate concept, study of examples) and integration of knowledge (comparison of some concepts within a class, tracing relations between the concepts, exploration of the related concepts). Based on ALD model, we suggest the following solution for the reconstruction of a interactive dictionaries into LDS: 1. Separation of dictionary knowledge structure from its context-dependent behaviors - a tools. It allows for creation of a unified Dictionary Player for various and distributed dictionary knowledge structures/objects. 2. To support cataloging of dictionary resources on the Net. Currently dictionary resources are cataloged by common used surrogates (Crimea, 1999) such as bibliographic descriptions in MARC format. Structure and parameters of those descriptions are intended for description of any resource, thus, a learner cannot determine specific parameters of dictionary objects, the depth of their description etc. 3. Dictionary Player is based on ALD model and supports learning “Wish-Trees” (see details about “Wish Tree” in (EDT, 1999)): agent role-play and plan-oriented autonomous learning of stable and right created knowledge structures/objects collections and related learning resources; decomposition/composition of autonomous learning content; structuring and ordering of that content and its cataloging into personal library/database; simulations of personal performance content and/or of missing features; multiple linked representations from different viewpoints of ALD model and performance content with various linked metadata collections for learner and/or agents. For all this, performance content (history, current, new] is more explicit for autonomous learner. 4.0 Programming for Learning-Centric Interactive Dictionaries An autonomous learning is based on ALD model and is executed by programs, which are also based on ALD. A simple example of a navigation program fragment: /* Example of common fragment; /* for [<+personal/broader context>] do [[ <+Right> <+Learning>] [<+Right> <+KnowledgeStructures/Objects>]] while [[<+Right> <+Time>] .AND. [<+Right> <+Place>& <+Cost>]] enddo /* /* (Where: “+” = one or more elements required; à [] - designation of container /* (WFC, 1996), (CSM, 2000); <+ personal/broader context > and <+Right> is Distributed /*Active Relationships (DARs, 1997), (CSM, 2000); /* /* /*EXAMPLE: some personal fragment /* /* [<Annotation>] = [<Goal2>] /* “Learn phraseological usage of a term in multilingual glossaries” /* /* [<Goal2>]= [[<Learning>][<Phraseologisms>] /* /* [<Annotation>] = [<Objective>] /* “Navigation /* in mono and multilingual glossaries” /* /* [<Objective1>] = do [<PerformNavigation>] /* while [<Glossary>] = [<Glossary1> .or. <Glossary2> ] /* enddo /* /* [<Objective2>] = do [<EvaluateNavigation>] /* while [<Glossary>] = [<Glossary1> .or. <Glossary2> ] /* enddo /* /* [<Plan1>] = [[<Objective1>][<Objective2>]] /* for [[<+MyWorkPlace>]=<Home>.or.<Mobil>]] do [<PerformPlan1>] while [[<InternalLinkOf SemanticStates>]= [<PhraseologicalLink>]] enddo /* /* [<PerformPlan>]=if [[<+CurrentEntry>]=[<+SemanticState>]] /* then [<NavigateInSemanticStates>] /* else [<GotoPlanEntry>] /* endif /* /* about SemanticStates see in (ShirSEM, 1999). /* ENDofEXAMPLE 5.0 Metadata Structures for Learning-Centric Dictionary Resources Dictionary environment in any of its behavior states has: <+Agent>, <+Role>, <+Action>, <+Information Resource> (UkrINTEI, 1999) and <+Relationships> (CSM, 2000). Agents transform Information Resources by the Actions. Agents may play different Roles in transformarions of the performance content. Agents and Information Resources combined by the Actions support the Roles. Examples of Agents are: persons, organisations, search instruments, dictionary authoring environments, agents of third-party such as brokerage systems. Examples of Information Resources are: a dictionary-style document, standards and/or base, WWW-pages. Examples of Actions are: originate, deposite, reformat, deliver. Examples of Roles see in Table 1. Metadata is a descriptive label, that can be used to index resources to make them easier and more convenient to find and use. Such labels are “ data about data” and are refered to as “metadata” (WFC, 1996). We use a more elaborated definition of metadata given by Lorcan Dempsew and Rachel Herry “Metadata is data associated with objects which relieves their potential users of having full knowledge of their existence or characteristics. It supports a variety of operations” (Dempsey, 1998). Therefore, <+Agent>, <+Role>, <+Action>, <+Information Resource> and <+Relationships> have data, which are associated with them and used for support of the autonomous learning. Definition/setting/using of those metadata is based on ALD Model. Dictionary metadata structures are described in (ShirMon, 1998), (HCI, 1999), (Crimea 99), (CSM, 2000). 6.0 Case Studies Ukrainian Lingua-Informational Fund of the National Academy of Sciences of Ukraine is working on the design of an electronic version of new generation of academic Ukrainian general purpose dictionaries (de facto dictionary standard in Ukraine) with a unified Dictionary Player. This product is oriented at the use in educational bodies, libraries, state departments and private companies and is supported by the leading language and learning centers of Ukraine. This project has been initiated by the Decree of the President of Ukraine “ On the development of the National dictionary bases” and is financed by the State innovation fund of Ukraine. STACCIS project, in which International Research and Training Center UNESCO/IIP has been involved, was focused on dissemination of information about telecommunication applications, tools and techniques through the network of so-called "information and demonstration centers". Besides a number of electronic catalogs comprising references to electronic publications, software, tools, services, distance courses and various collections, it will include as an extension a Basic Internet Glossary with a unified Dictionary Player for novices in telecommunications to ease search of relevant topics and facilitate professionally-oriented learning. The above mentioned unified Dictionary Player based on ALD Model will support the following functions: * definition/use of personal roles, goals, plans as learning objects for autonomous learning of our dictionary collections; * navigation in collections of the learning objects and additional functionality to support navigation in a broader space; * structuring and ordering of personal performance (history, current, new) content and its cataloging with an annotations in the personal library/database; * simulation of personal performance content and of missing features; * evaluation of behaviors; * management of self-learning; * cognitive support for the mentioned functions. 7. 0 Conclusion Existing interactive dictionary model ab inito is a closed system. Autonomous learners as users of this closed system cannot resolve their problems in an open LVE. A fundamental question of the approach to a interactive dictionary as a closed system is: “How a dictionary may be linked to other dictionaries or third-party resources situated outside of the closed environment?” In a real world, interaction of interactive dictionaries with an “external world” is not supported. An autonomous learner needs more resources than may exist in any “closed world” of interactive dictionary. Therefore, an ideal solution for autonomous learning: dictionary model is an open system. REFERENCES 1. (Colazzo, 1996) Colazzo L., Costantino M. (1996) Using multi-user hypertextual glossaries in multimedia-based design. Proceedings of EuroAIED – European Conference on Artificial Intelligence in Education. Lisbon, Portugal. 2. (ComNEd'99) Sinitsa, K. & Manako A. (1999) Extending glossary role in a virtual learning environment. In Downes T. & Watson D. (Eds.): Communications and Networking in Education: Learning in a Networked Society, (Post-conference book of ComNEd'99, Aulanko, Finland, June 13-18, 1999), pp.321-327/ 3. (Crimea, 1999) Shirokov, V,, Manako, V., Sinitsa, K. & Manako, A. (1999). Adaptive digital library support for creation and exploitation of the next generation of Ukrainian mono- and multilingual dictionary resources. In Proceedings of the 6 International Conference “Crimea 99” on Libraries and Associations in the Transient World: New Technologies and New Forms of Cooperation, vol. 1. - Moskow: GPNTB of Russia, 1999, pp. 268-270. 4. (CSM, 2000). Широков В.А., Манако А.Ф., Манако В.В., Синиця К.М. Вступ до архiтектури цифрових словникiв. Journal of “Control Systems and Machines”, 2000, No. 1, pp. 5. (DARs, 1997) Daniel Jr., Ron and Carl Lagoze, "DistributedActive Relationships in the Warwick Framework", Proceedings of the 1997IEEE Metadata Conference, September, 1997, 6. (Dempsey, 1998) Dempsey, L. and Heery, R. Metadata: Current view of practice and ussues. Journal of Documentation, 1998, 54(2), pp. 145-172. 7. (DICsit, 1998) Рубан В.Я., В’юн В.І., Кузьменко Г.Є., Морозов А.О., Широков В.А., Пещак М.М., Шевченко І.В. Українськоросійсько-англійський термінологічний словник з ситуаційного управління. //Математические машины и системы. № 1, 1998 , С.128–173. 8. (EC’95g) European Commission, The green paper on innovation, December 1995. 9. (EDT, 1999) Sinitsa, K. & Manako A. (1999). Interactive Dictionary as an Information Wish-Maker. Educational Technology, vol.39 (8), pp.22-25. 10. (HCI, 1999) K. Sinitsa, A. Manako, Interactive Dictionary in a context of learning // Proceedings of 8th International conference on Human-Computer Interaction: Communications, Cooperation and Application Design, Volume 2 / edited by Hans-Jörg Bullinger and Jürgen Ziegler / Lawrence Erlbaum Associate, Publishers, London / ISBN 0-8058-3392-7, 22 – 26 of August, 1999, Munich, Germany. - P.662666. 11. (ModSUM, 1996) Широков В.А., Пещак М.М.. Структурна модель реєстрової частини Словника української мови. // “Національна архівна інформаційна система “Архівна та рукописна Україніка і комп'ютеризація архівної справи”, Зб. наук. праць. вип. 1: “Інформатизація архівної справи в Україні: сучасний стан та перспективи". – К., 1996. – С. 154-174. 12. (MOV, 1996) Русанівський В.М., Тараненко О.О., Широков В.А. Теоретико-лінгвістичні засади та інформаційно-комп'ютерне забезпечення україномовних лінгвістичних інтелектуальних систем // Мовознавство. – 1996. – № 4-5. – С. 3-8. 13. (OCLC) CNI/OCLC Workshop on Metadata for Networked Images - Executive Summary, http://www.oclc.org:5046/research/dublin_core/summary.html" 14. ( POINTER) Final report of POINTER http://www.mcs.surrey.ac.uk/Research/CS/AI/pointer/report/section4.html EC Project. Terminology resources. 15. (ShirMMS, 1999) Широков В.А. Строение лексикографических систем. //Математические машины и системы. –К., № 3, 1999, С.17–39. 16. (ShirMON, 1998) Широков В.А. Iнформацiйна теорiя лексикографiчних систем. - К.: Довiра, 1998, - 331 с. 17. (ShirSEM, 1999) Широков В.А. Лексикографічне представлення семантичних станів. //Математические машины и системы. –К., № 2, 1999. С.11–21. 18. (Sinitsa&Mizoguchi, 1998) Sinitsa, K, Mizoguchi, R. A Structured Hyper-Glossary as a Cognitive Tool for Computer-Based Learning, Proceedings of CATE-98, International Conference on Advanced Technologies in Education, Mexico, 1998. 19. (UkrINTEI, 1999) Shirokov V,, Manako V., Sinitsa, K. & Manako A. (1999). Framework for Ukrainian Digital Dictionary Resources. Proc. of the 6 International научно-практическая Conference “Проблемы создания, интеграции и использования научнотехнической информайии на современном этапе”, Kiev, Ukraine, December 16-17, 1999, Kiev:UkrINTEI, pp. 20-21. 20. (UkrPROG98) Shirokov, V (1998). Технолингвистика: модели данных и проблемы программирования. УкрПРОГ’98, pp.451-461. 21. (WFC, 1996) Lagoze, Carl and Lynch, Clifford and Daniel, Ron, Jr. June, 1996. The Warwick Framework: A Container Architecture for Aggregating Sets of Metadata. Cornell Computer Science Technical Report TR96-1593.