Guenther Goerz

The WissKI system provides a framework for ontology-based science communication and cultural heritage docu-mentation. In many cases, the documentation consists of semi-structured data records with free text fields. Most references in the texts comprise of person and place names, as well as time specifications. We present the WissKI tools for semantic annotation using controlled vocabularies and formal ontologies derived from CIDOC Conceptual Reference Model (CRM). Current research deals with the annotations as building blocks for event recognition. Finally, we outline how the CRM helps to build bridges between documentation in different scientific disciplines.

The Epigraphische Datenbank Erlangen-Nurnberg / Epigraphic Database ErlangenNurnberg (EDEN) is an evolving online database of ancient Greek inscriptions from cities of today’s western Turkey: Metropolis (in Ionia), Magnesia ad Maeandrum and Apollonia ad Rhyndacum. The development of EDEN started 2012 at the University of Erlangen-Nurnberg in close collaboration between ancient historians, archaeologists and computer scientists. Contentwise, the database is maintained by the Chair of Ancient History. The Digital Humanities Research Group is responsible for the technical framework. The cooperation assures the rapid evolution of the database despite the tight budget. At the time of writing, the database contains nearly 600 Greek an Latin inscriptions of Hellenistic and Roman emperial period enriched with metadata.

This paper reports on on-going research implementing the FRBRoo (Objectoriented Definition and Mapping to Functional Requirements for Bibliographic Records) in the description logic language OWL-DL. This implementation is called Erlangen FRBRoo (EFRBRoo). The goal to be fully compatible to Erlangen CRM / OWL (ECRM) is achieved by applying the same implementation patterns of ECRM wherever possible. In order to preserve the FRBRoo specifications at the same time EFRBRoo keeps also as close as possible to the text of the FRBRoo definition. The EFRBRoo is made available online. To evaluate and improve the quality of the EFRBRoo implementation, the software development method of static unit testing is applied: simplified test cases have been designed, which consist of representative datasets in OWL-DL, SPARQL queries and respective results. In addition, validators (RDF, OWL) and frameworks (Protégé, Sesame, and WissKI) have been used for validating formal correctness and class consistenc...

This paper places the concept of ‘common sense geography’ as developed by the members of Topoi research group C-5 within the context of ancient geographical literature. For the first time, a consistent model of arranging and classifying Greek and Roman geographical texts from a historical perspective is presented.

CIDOC’s CRM has been introduced as a formal reference ontology with a particular focus on cultural heritage documentation. Meanwhile acknowledged as an ISO standard (21127:2006), one of its main goals is to facilitate interoperability between data and database schemata. In order to take full advantage of CRM’s benefits, an implementation in a modern knowledge representation language has been a desideratum for quite a while. Suitable candidates for very expressive, albeit decidable knowledge representation languages are Description Logics. A dialect of the Web Ontology Language, OWL-DL, based on XML and RDF/RDFS(FA) within the Semantic Web language hierarchy, is equivalent to a very expressive description logic. To satisfy the needs arising from practical applications, we have developed an implementation of the most recent CRM version in OWL-DL. We will discuss problems of formalization of the scope notes by a few typical examples, and, furthermore, point out some problems in the des...

In this paper we present the WissKI System, an open source Virtual Research Environment and Content Management System for Cultural Heritage. It promotes semantic enrichment of data on the basis of OWL / RDF using the ontology CIDOC CRM / ISO 21127. The data is rendered in a Wikipedia-like fashion, combining textual, visual and structured information (info boxes) for a documented object on one page. Likewise, data can be acquired using field-based forms and semi-automatically annotated free text, resembling the most common traditional modes of documentation in the cultural heritage domain. This retains a user-friendly visualisation while at the same time providing detailed RDF triple data for automatic processing and data exchange.

In a collaborative effort between IT and museum experts, the research project “Objekte im Netz” aims at developing a joint digitization strategy for the digital documentation of heterogeneous university collections by applying semantic web technologies. The University of Erlangen-Nuremberg owns more than 20 different scientific collections. So far, each collection has documented and stored its object data in different ways, thus making it impossible to interlink the information. The data has neither been stored in a sustainable way nor has it been reusable. In summary, the stored information is not utilized to its full potential. This is a typical scenario regarding scientific collections at universities in German-speaking countries. In order to allow consistent documentation for all collections, a data model based on ICOM’s CIDOC Conceptual Reference Model is currently in development. The model uses sample data of six representative collections of the University of ErlangenNurember...

Behaims „Erdapfel“ von 1492 ist der alteste erhaltene Erdglobus. Er ist ein fruhes Meisterwerk, das von einer Reihe fruher wissenschaftlicher und technischer Innovationen zeugt. Heute gehort er zu den prominentesten Exponaten des Germanischen National-Museums in Nurnberg. Sein Kartenbild ist primar ptolemaisch, es enthalt aber auch Elemente der mittelalterlichen Universalkartographie und von Portulanen. Seine umfangreiche und kunstvolle Ausstattung umfasst uber 100 Miniaturen und sechzig Fahnen und Wappen, mehr als 2000 Ortsnamen und uber 50 lange Inschriften. Der Behaim-Globus ist unter den erhaltenen kartographischen Werken nahezu einzigartig dadurch, dass in ihm verschiedene Traditionen der spatmittelalterlichen Kartographie vereinigt sind. Wir berichten uber ein laufendes Forschungsprojekt, das eine digitale und eine gedruckte Edition des Globus zum Ziel hat. Im Jahr 2011 wurden hochaufgeloste digitale Fotografien aufgenommen und ein neues 3D-Modell erstellt. Daruber hinaus gibt...

Bibliotheca Hertziana’s Biondo research group questions an epistemology of spaces and their changes in the early modern history. At focus are relations between historical maps and texts aiming to explore the historical understanding of spaces and the knowledge associated with it. We take up approaches from cognitive science and computational linguistics arguing that cognitive maps depict culture-specific spatial knowledge and practices. Our interdisciplinary project combines cognitive-semantic parameters such as toponyms, landmarks, spatial frames of reference, geometric relations, gestalt principles and different perspectives with computational and cognitive linguistic analysis. Using new text and map markup and corpus-specific quantitative methods, historical geographical texts are processed and reinterpreted. Long-term research questions are: Which forms of knowledge represent spatial relations? How can spatial transformation processes be represented and analyzed? What is the con...

Abstract. We present an interdisciplinary approach to implicit knowl-edge of spatial cognition in common sense geography. Structures such as, e.g., distance, scale, topology, trajectory and frames of reference are believed to be encoded as mental maps. Furthermore, we refer to com-mon sense as ‘naive ’ perception and descriptions of space and the use of ‘intuitive ’ arguments in geographical contexts. The empirical data sets comprise of ancient written texts from different periods and sources. Our methodology combines annotating and parsing techniques based on corpus data analysis as a (semi-) automated analysis and cognitive pa-rameters applied in cognitive linguistics. These parameters are based on gestalt-psychological principles such as figure-ground asymmetries. We argue that the survey of ancient texts provides further insights whether there are basic epistemological expressions of spatial orientation that might be candidates for universals. As a first example, we investigate

We considertheoreticalaspectsfrom linguistics andlogic for contributionsto the problem of systemscalability. For the linguistic part, we claim that a “pragmatics-first”view on rational interaction providesanappropriateframework for flexible and scalabledialoguemodelling. In particular, theplan-basedapproachoffersthemeansto conducttask-or goal-orienteddialogueswhich aimataccomplishingconcretetasks.It enables cooperati ve responsebehaviour andtheability for negotiation. For thereasoningpart,i.e.knowledgerepresentationandinferencefor theinterpretationof dialogueas well as for planningto satisfy user goalsin theapplicationdomain,wearguefor a computationalogic framework. Thereis no doubt that a minimal prerequisite for scalablesystemsis that they have a modular structure.We claim that a clearfunctional separationbetweenthelanguagemodel,thedialoguemodel,andthedomainmodelprovides a sufficientconditionto addressscalability. 1 GeneralAssumptions Our generalgoalis to build dialoguesystemsfor rational interaction.Whatwewantto achieveis thesatisfactionof usergoalsin agiven(ideallyopen)domainby conducting spokendialogueswhereit shouldbepossiblein principle to augmentthemby otherformsof multi-modalinteraction like gesturesor the selectionof itemsfrom a menu on a screen. Interactionsare called “rational” because wewantto applyrationalityprinciples(at theknowledge representationlevel) to optimallyselectappropriatecommunicati ve actions. We assumethat the satisfaction of usergoalswithin the thematicframework of a particular applicationdomainis to beachievedwith thehelpof a dialoguesystemproperin cooperationwith a technical applicationwhich we alsocall the “domainproblem solver”. Sucha technicalapplicationcanbeaninformation or reservation system,a systemfor controlling certaindevices,etc. In suchsettings,(Allen etal.,2001)characterize“practical systems”astask-or goal-orienteddialoguesystems, wherethe dialogueis focussedon accomplishinga concrete task. They claim that the conversationalcompetencerequiredfor practicaldialoguesis significantlysimpler to achieve thangeneralhumanconversationalcompetence.We areconvincedthat this distinctionis useful becauseit providesa realisticstartingpoint andbecause the latter – generalhumanconversationalcompetence– is hard to define. Hence,we begin with a certainwelldefinedconversational competenceandwith increasingly complex requirementsfrom theapplicationwe cantry to augmentit, i.e., we attemptto proceedin an incremental fashion. But probablythis will not be easy;we cannot expect that this can be donein the way of a linear progress:Theremaybe fractionswherecompletelydifferentrequirementscomein which cannotbe integrated seamlessly . Anothergeneralunderlyingassumptionis that for the interpretationof dialoguewe insist on a clear commitment to a (computational)logic framework. Of course, humansact incoherentlyand even inconsistently, and commonsensereasoningcanonly to a certainextentbe understoodin termsof logic, but we areconvincedthat a coherentand consistentrational reconstructionis the bestwe can do about it. Sucha constructi ve perspectivehastheadvantageof enablingusto begin with awell understoodframework for knowledgerepresentationand reasoninguponwhich we canattemptto build rule systemsfor still idealized,but morerealisticpatternsof argumentationin specificdomains.Webelieve thatthereis a potentialto succeedin a varietyof prevailingly instrumentalizedcontextsasit is thecasewith technicalapplications– thatwill bediscussedin moredetailbelow – or, to take upanotherexample,in forensicargumentation. 2 Remarks on Scalability The question of scalability of natural languageand speechdialoguesystemshasbeenan issuefor quite a long time. And it becomesmoreandmorepressingwith therapidly increasingprocessingpowerandmemorysize of modernhardwareand the progressin softwareengineeringtechniques. Beforeweaddressthedimensionsof scalabilityandtheproblemsconnectedwith it, let ushave a look at historyto sharpenour awareness. 2.1 In the Beginning: The Structural Approach For severaldecades, naturallanguageandspeechsystems for informationdialogueshavebeendevelopedonthebasisof anapproachwhich led from experimentalsystems to a commerciallyavailable technology. Although successfulin many domains,we will argue that thosesystemshave severe limitations which are inherentto the underlyingso-calledstructuralapproach. The goal addressedby this first systemgenerationis to accomplishan informationtask. By conductingmore or lessstrictly guideddialoguesthey aim at supplying a user with a specific information in a given domain, e.g.train connections, traffic jams,cinemaprogrammes, stockexchangedata,weatherinformation,etc. in asfew dialogueturnsaspossible.Thetechnicalapplicationthey arecooperatingwith is usuallya staticdatabasesystem; in somecasesit hasbeenextendedwith an application server that might be taking ordersor reservations. The basictechniquethesesystemsuseis theextractionof parametervaluesfor a given schemafrom…

this paper we describe the overall system and the approach towards robust dialogue processing. The system utilizes spoken input and output and is completely integrated operating on a single general purpose workstation with a total response time of typically 3 seconds. This fast system reaction is important for an user acceptance. Since the acoustic recognizer still returns wrongly recognized utterances, further processing stages have to cope with robust interpretation techniques. 2 System

The goal of many hypermedia systems, which are a synthesis of hypertext and mul-timedia systems, is to open complex information spaces through the use of associative networks with nodes containing texts, images, and acoustic signals. The central issue of this paper is to argue that the power of such information systems can be considerably enhanced by the integration of a knowledge representation component. We illustrate this thesis with an example hypermedia information system about the oldest existing globe of the earth whose creation was initiated by Martin Behaim in 1492. Displaying the pre-columbian image of the earth, the Behaim globe is a masterpiece of art of the European Renaissance. Today it is one of the most attractive objects in the collections of the Germanisches Nationalmuseum in Nuremberg. Our hypermedia information system consists of two parts, a graphical visualization component which can produce images of the globe from any desired view, and a hypermedia component which provides information about the globe, in particular diierent readings of the texts written on it, various comments on the texts and miniatures on the globe, images of contemporary globes and maps, a complete record of scientiic publications about the object, and cultural and historical background information. Particular attentention is paid to the broad variety of relations between images and texts.

ABSTRACT We describe progress made within the FORK project, whose goals are the implementation of a primarily object-oriented knowledge representation system and its application to the design and fault diagnosis of technical systems. Whereas the kernel of the FORK representation system is completely object-oriented, the system as a whole is supposed to integrate a variety of different programming styles. In the following, an extension for rule-oriented programming is described, which raises the descriptive power of the FORK system beyond that of LOOPS. As an application of the rule-oriented component, a constraint language has been implemented which plays an important rule in our approach to the design and fault diagnosis of technical systems. The next steps in the FORK project will include the development of a general logical framework, comprising a logical reconstruction of object-centered representations, retrieval of complex descriptions by unification, and deductions on structured objects. The problem of non-monotonicity will be dealt with on the meta level by a module similar to DeKleer’s ATMS. Further progress shall be achieved by concentrating on a general treatment of the problem of time in modelling technical systems which is to our opinion one of the most important issues.

Integration of new utterances into context is a central task in any model for rational (human-machine) dialogues in natural language. In this paper, a pragmatics-flrst approach to specifying the meaning of utterances in terms of plans is presented. These plans are computed dur- ing a dialogue on the basis of information about the current situation that is updated continually. New

The goal of most hypermedia systems is to open up complex information spaces through the use of as- sociative networks of nodes containing texts, images, and acoustic signals. The power of such information systems can be considerably enhanced by the integration of knowledge representation. We illustrate this claim with an example hypermedia information system about the oldest existing globe of the earth whose creation was initiated by Martin Behaim in 1492. Displaying the pre-columbian image of the earth, the Behaim globe is a cartographicalmasterpiece of German humanismin the Renaissanceage; nowadays the globe is exhibited in the GermanischesNationalmuseum in Nuremberg. From the user viewpoint, our hypermedia information system offers two paths for information access, one through a graphicalvisualization component which can produce imagesof the globe from any desired viewpoint, and a graphical query interface to a hypermedia network which provides structured informa- tion unitsconsistingof texts andimages. Theseunitscontain descriptionsof thedifferent globeregionsand their map image, various readings of its numerous inscriptions, comments on the texts and miniatures on the globe, images of contemporary globes and maps, a comprehensive record — in part on-line accessible — of important scientific publications on the globe, and cultural and historical background information. Particular attentention is paid to the broad variety of relations between images and texts. For the information system, a new distributed software architecture based on WWW technology has been specified and implemented. The server component consists of the system kernel and several server components, to which the kernel dispatches upon request. Among the components are an object-oriented database system with a specific query component (via CLOS, the Common LISP Object System and Al- legro/Object Store), an image database and preprocessing component, and a component for generating hypertext (HTML) pages from database contents. The object-oriented database system holds all data ob- jects describing the globe, its parts, and background information, as well as descriptors for all kinds of media data. Immutable image data and text files (scanned documents) are kept separately, but are accessi- ble only through these descriptors.

In parsing natural language, incremental semantics composition is one of the most prominent issues. In the past, numerous approaches have been developed for assigning meaning to noun and verbal phrases and their complements and modiers. Often, their inferential power is too low for practical applications or the expressiveness of the representation language leads to intractable inference procedures. As an answer to these problems, we discuss an approach that relies on Description Logics for handling this class of semantics construction. We show how a semantic knowledge base can be setup. We exploit the equivalence between Dis- course Representation Structures limited to the expressiveness of ALC and ABoxes for validating DRS with respect to a given knowledge base.

In this paper we describe the linguistic processor of a spoken dialogue system. The parser receives a word graph from the recognition module as its input. Its task is to find the best path through the graph. If no complete solution can be found, a robust mecha- nism for selecting multiple partial results is applied. We show how the information content rate of the re- sults can be improved if the selection is based on an integrated quality score combining word recognition scores and context-dependent semantic predictions. Results of parsing word graphs with and without pre- dictions are reported.