The X-COSIM Integration Framework for a Seamless Semantic Desktop

The X-COSIM Integration Framework for a Seamless Semantic Desktop Thomas Franz, Steffen Staab, Richard Arndt University of Koblenz-Landau Universitaetsstr. 1 56070 Koblenz, Germany {franz, staab, rarndt}@uni-koblenz.de

ABSTRACT In this paper, we present X-COSIM, a framework for cross context semantic information management consisting of the ontology X-COSIMO and the application programming interface X-COSIMA. X-COSIM provides what is lacking in current semantic desktops: A consistent seamless integration of a conceptually broad reference model and context dependent conceptualizations as required by applications dedicated to supporting complex work processes. We present the design of XCOSIMO and illustrate the application of X-COSIM in a scenario that involves multiple desktop applications.

Categories and Subject Descriptors H.4.1 [Information Systems Applications]: Office Automation

General Terms Design, Human Factors

1.

INTRODUCTION

The area of personal information management (PIM) is concerned with supporting personal work processes that are commonly complex (i.e. consisting of subtasks), information intensive, ad-hoc, and context-spanning [8]. Within such work processes information is reused, manipulated, and retrieved in different contexts. The semantic desktop [3] improves support for personal work processes by employing semantic data models in order to let PIM applications share data. An open challenge in semantic desktop research is how to mediate between different contextualized views required by individual PIM applications that support different working contexts. For example, email clients build upon representations of senders, recipients, and attachments Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. K-CAP’07, October 28–31, 2007, Whistler, British Columbia, Canada.

while file browsers are based on representations of system users and files. The information underlying these representations, however, overlaps, e.g. a person can be the sender of an email and the holder of an operating system account. In this paper, we present X-COSIM1 , our approach to that challenge. X-COSIM stands for cross-context semantic information management and provides a framework consisting of the ontology X-COSIMO and the application programming interface X-COSIMA. X-COSIM enables knowledge work by supporting both information reuse across contexts but also contextualized, task specific views onto information. It is based upon a principled approach towards providing conceptual models based on a foundational ontology, i.e. DOLCE [10] which provides a large conceptual denominator for a broad scope of complex task support and hence for a large variety of PIM applications. Using a design pattern for formalizing context, contextual views can be represented within the reference ontology X-COSIMO while maintaining the consistency of underlying information. In combination with X-COSIMA, X-COSIMO establishes a semantic desktop architecture for building PIM applications that leverage complex personal work processes by supporting information linkage and reuse across work contexts.

2.

ISSUES WITH CURRENT DESKTOPS

In this section, we discuss several issues with current desktop implementations when it comes to complex task support. The discussion is focussed around the following generic tasks that commonly interplay and overlap when performing a complex task: 1. Classifying Information: e.g. structuring and summarizing information (Sect. 2.1) 2. Communicating: e.g. sharing, arguing, explaining (Sect. 2.2) 3. Retrieving and Reusing Information: e.g. modifying, filing and browsing information (Sect. 2.3)

1

http://isweb.uni-koblenz.de/Research/x-cosim

2.1

Issues in Information Classification

Currently, PIM tools prevent a simple reuse of classification metadata, i.e. a sharing of the classifications made. As an example, web browsers, file managers, and email clients usually provide folder hierarchies to classify content. These hierarchies, however, cannot easily be shared and reused among current PIM applications.

2.2

Issues in Communication

There exist various computer based modes of communication with varying suitability for different communication tasks [5]. Communication clients, however, are commonly built to support only one mode (e.g. email) so that conversations across multiple modes are penalized by reduced traceability of conversation threads. It is also known that modes of communication are not exclusively used as pure communication means [9] but exploited for various purposes such as task management (e.g. using emails as reminder), contact management, or simply as a medium for data transfer (e.g. file transfer over instant messaging protocols or using email attachments)[12]. Accordingly, a message in the sense of the protocol can play multiple roles depending on the context or user task, e.g. it can be a textual message (context communication), an event description (context calendaring), a textual annotation of an attachment (context information browsing), or a request for work (context task management).

2.3

Issues in Information Reuse & Retrieval

Current desktops provide file browsers for filing and retrieving information. However, when using a file browser, metadata from contexts other than file browsing is not available, e.g. for a file shown in the file browser we cannot inspect the email to which it was attached, find out about the web page from which it was downloaded, view the event description for which it was created, or the task it is associated with. While such contextual information would be beneficial in various tasks it is not provided by today’s desktops. Filing mechanisms of desktops do not provide information about changes and origin of information, e.g. when saving a file under a new name, no relation to the original file is maintained. Similarly, when several files are merged into a single one, no linkage to original files is maintained, e.g. when creating a presentation that is composed of several images and text documents.

2.4

Analysis

Analysing the issues described before, we conclude that current PIM tools make cross-context information management difficult as they do not support information reuse across tools, do not maintain information linkage and thus cannot preserve metadata from other contexts. Redundant maintenance efforts and increased cognitive effort are the result. These shortcomings of current desktops are an implication of the following design decisions:

1. Implicit Models: Current PIM applications are often based on implicit data models, i.e. how information is modeled, accessed, and what certain information represents is expressed only within the application’s source code. Accordingly, implicit models prevent access to information from other PIM applications which prevents information reuse and linkage. 2. Data Formats: Some PIM applications feature the import and export of information in a standardized or documented data format, e.g. as comma separated values (CSV) or expressed in XML. These formats, however, only define a structure but no content semantics, i.e. they do not explicate what some information represents so that data cannot be interpreted by machines. 3. Narrow Conceptualizations: The models employed by current PIM tools represent context dependent and thus narrow views onto the information they deal with. The underlying conceptualizations conflict with conceptualizations of the same information given by tools that support another context. For example, in the file system context a person is conceptualized as owner of an operating system account to which ownership, access and execution rights for files are associated. While in the context of task management, a person may be modelled as someone responsible for the production of certain output, characterized by particular skills, availability times, and authority. Though the two conceptualizations of a person may overlap, they cannot easily be aligned automatically. Moreover, models may contain ambiguous conceptualizations of similar but different aspects so that mapping models of different PIM tools becomes difficult, complex and error-prone. The observation of such shortcomings led to the idea of a semantic desktop that mediates between different applications with explicit semantic data models in order to let them share data seamlessly. First proponents like Haystack [7] and Gnowsis [11] turned out to be very intriguing as they solve issues resulting from implicit data models and missing semantics in data formats (issue 1, 2). The overall task of providing a middleware that mediates between narrow conceptualizations as employed by different PIM applications (issue 3), however, proved to be extremely difficult for several reasons (see also Sect. 7): a) One-to-one mappings do not scale: Approaches like Haystack assume that each application has a unique semantic data model. Supporting information exchange between any two semantic applications requires a one-to-one adapter between the two corresponding data models. Providing such adapters becomes more and more difficult as the number of PIM applications grows. b) Complex tasks are not supported : Users view, ma-

nipulate and retrieve information across contexts in complex tasks. Individual PIM applications tend to be specific for (nearly) atomic tasks. This implies that their data models will be of narrow scope and not be able to capture all the information required for a complex task. When crucial data is not propagated, support of the overall task will break easily. c) Lack of development methodology for PIM ontologies: Finally, existing PIM ontologies have been developed in a rather ad hoc way further hampering their extension as well a subsequent mediation between applications. In the following sections, we explain how we tackle the issues enumerated in this section, in particular the issues a) to c) that represent shortcomings of current semantic desktops.

3.

X-COSIM ARCHITECTURE

The X-COSIM architecture is based on the analysis of shortcomings found in current semantic desktop prototypes (Sect. 2). As a result, the architecture consists of three conceptual layers (cf. Fig. 1): To tackle issues of information linkage and reuse, the bottom layer provides X-COSIMO as the reference ontology that is a sound model, conceptually broad enough to enable the consistent capture of any information relevant for PIM. The middle layer is introduced to provide components for the mapping between the bottom layer and the top layer. It addresses the issue of mapping between different data formats characterized by narrow conceptualizations (top layer) and the single reference ontology (bottom layer). The programming interface XCOSIMA is one component of the middle layer. The top layer is the application layer including the domain specific conceptualizations as employed by individual PIM applications.

Figure 1: X-COSIM Architecture The following sections are aligned to the layers of the X-COSIM architecture, we present X-COSIMO, the reference ontology in Sect. 4, describe the mapping layer including X-COSIMA in Sect. 5, and illustrate an application scenario in Sect. 6.

4.

X-COSIMO

X-COSIMO provides the reference ontology within the X-COSIM architecture. It is divided into modules that represent different aspects of information: The

information realization module (Sect. 4.2) addresses shortcomings related to information reuse and retrieval (Sect. 2.3) by making explicit the relation between information and its realization by different information systems. Computer mediated communication is modelled by the communication module (Sect. 4.3) that abstracts from particular communication protocols and channels. It provides metadata about communication aspects for reuse in different work contexts. It also provides a unified view onto communication to tackle issues of communication applications (Sect. 2.2). The classification module (Sect. 4.4) provides a model for the annotation of information to enable the reuse and sharing of user-defined classifications across work contexts, e.g. to reuse a concept that classifies a web page in the context of web browsing for the classification of an email in the communication context (Sect. 2.1). The decomposition module (Sect. 4.5) enables the segmentation of information enabling to link, reuse and annotate particular parts of information, e.g. to represent compound documents such as presentation documents that are composed of several other documents (Sect. 2.3). Before we explain the modules of X-COSIMO in detail, Sect. 4.1 introduces the methodology for the development of the modules.

4.1

A Reference Model with Context Support

Due to the diversity of information relevant for PIM (e.g. information about tasks, people, conversations, files, meetings), a model that captures such information needs to provide a broad conceptual scope that enables a consistent representation of that information. Therefore, the development of X-COSIMO is based on a foundational ontology, namely the DOLCE library of foundational ontologies [10]. From the alignment with DOLCE, X-COSIMO inherits its rich axiomatization and broad scope which spans all top-level categories. New conceptualizations as required by new applications are required to align their concept definitions to the top-level concepts. Thus, one may avoid the inclusion of unnecessary and inconsistent modelling artifacts which may cause incompatibility and prevent information reuse. Consequently, the choice of building upon a foundational ontology is also a means to ensure consistent evolution and extension of X-COSIMO. Naturally, models built to represent contextualized information are of narrow scope preventing information reuse and linkage (cf. Sect. 2). However, such context dependent conceptualizations are needed to support application developers, i.e. enabling developers of communication tools to deal with concepts such as message, sender and address instead of more abstract and complex concepts as defined by a foundational ontology. Accordingly, the requirements for context-dependent conceptualizations and information reuse impose two conflicting challenges: 1. Consistent representation of arbitrary PIM rele-

vant information under a single reference ontology 2. Integration of potentially inconsistent context dependent conceptualizations of the same information As a solution to these challenges the Descriptions & Situations (DnS) design pattern [4] is employed for the development of X-COSIMO. The DnS pattern builds upon a formalization of descriptions which represent context. Descriptions define Parameters, Roles and Courses that are used to classify DOLCE ground entities within a Situation. Situations satisfy Descriptions, if at least all ground entities that exist within the Situation are classified according to the Description, e.g. a particular conversation corresponds to a Situation that satisfies the rules defined by a communication theory. Thus, the DnS pattern can be employed to represent different aspects of information reconciled under the common conceptual model of the DOLCE ontology. XCOSIMO modules described in the following sections represent core aspects of information as relevant in personal information management.

4.2

Information Realization

The desktop and associated applications provide access to information as realized by information systems, e.g. information realized by a file on the file system, by an HTML page on a web server, or by a text field in a database. As discussed in Sect. 2, current desktop applications prevent cross-context information utilization partly due to their use of implicit models and proprietary formats. The module for information realization can be used to explicate how information is accessed and serialized by semantic metadata. Any desktop application can interpret the realization metadata for serving the realized information to the user. Moreover, the module builds upon a conceptualization of information that is independent of its realization. Thus, the realization module can be used to consistently represent different realizations for the same information while maintaining linkage between them. File type conversions and web downloads are examples that result in such multiple realizations, e.g. the same information may be realized by an OpenOffice document available in the OpenDocument format but also available in the portable document format (PDF) (for instance after using some “export to PDF” functionality). An image downloaded from a web page is both realized on a local file system as well as a web accessible resource on a web server. The realization module is implemented according to the DnS design pattern by modeling information realization as a situation of a Digital Realization (cf. Fig. 2). Such a situation satisfies a Realization Description that is based on the assumption that information realization requires the identification of data, a protocol that specifies data access, and a data format that specifies the serialization of data. Accordingly, the description of a realization defines i) an Identifier, ii) Protocol, iii)

Format, and iv) Realizer role. The first is played-by a Particular that realizes the identifier, e.g. the string representation of a URI or of a database key. The second role is played-by a description that expresses an access protocol specification, e.g. the internet message access protocol. The Format role is played-by a description that expresses a serialization format such as the rich text format (RTF). The fourth role is intended to be played-by a further Particular that holds the actual data, e.g. an RDF literal or a large binary object. While Fig. 2 shows only a few of the protocol and format descriptions available, the set of protocols and formats is not assumed to be closed and the module is extensible towards further protocol and format specifications. As an example, the information provided by a research paper is represented by an Information Object. The realization of that research paper (e.g. as a PDF document on a web server) is expressed by a Digital Realization that realizes that Information Object. The Digital Realization satisfies a Realization Description which defines the roles Protocol, Identifier, and Format. These are played-by a description of a protocol (HTTP) and format (PDF) as well as a Particular that plays the Identifier role storing the URL of the document.

4.3

Communication

The aim of the communication module is i) to enable a unified view onto communication and ii) to represent information dealt with in the context of communication for reuse in further contexts while maintaining information linkage. For achieving the first, the module provides a conceptual view of communication as developed by Jakobson [6] who defined the concept of a Message that is about something which he calls Context, that is transmitted via a Contact — a connection between the Addresser and Addressee — and that is expressed within a Code. Such a model generalizes communication so that arbitrary communication modes such as chat, phone and email can be represented consistently. For achieving the second goal, the module is implemented in the DnS design pattern as a Communication Description (cf. Fig. 3) that represents the communication model indicated before: The Communication Description defines the roles Addresser, Addressee, Contact, and Message as common roles for any kind of communication defined as by Jakobson. Jakobson’s concept of Context and Code are not defined by the description as the classification module (Sect. 4.4) and decomposition module (Sect. 4.5) express these aspects in a more general way that is applicable also to other contexts than communication. Additionally, a Communication Description also defines the roles Address and DispatchTime which are played-by Information Objects that represent address information of addressers and addressees as well as temporal information about the time at which a message was sent. Further DOLCE Endurants play the remaining roles: Agents play the

Figure 2: Information Realization Module (Classes defined by DOLCE in Gray)

Figure 3: Communication Module role of addressers and addressees, while descriptions of protocols play the role of Contact, e.g. a description of the simple mail transfer protocol plays the role of contact for email communication. As already indicated by Fig. 2, the Message role is played by an Information Object that is separated from its realization that is accessible by a Digital Realization. Next to communication roles, the Communication Description also defines a Communication Course which sequences Communication Events. A communication event can be a conversation that spans several exchanges of messages. Each role player such as an addresser and a message is a participant-in a Communication Event. While the Communication Description represents the abstract model of Jakobson, it subsumes more specific descriptions of particular communication modes, e.g. representing email communication or instant message communication. Descriptions of such communication modes define more specific roles than the abstract Communication Description, e.g. the Email Description defines the role of an Attachment and of an Email Protocol as more specific roles of a Message and a Contact. Applying the communication module for representing communication information, it can be reused in further contexts, classified according to any classification scheme and linked with further information. Moreover, disadvantages from the separation of different communication modes are alleviated as messages of any mode are represented by a common conceptual model for communication that results in a unified view onto messaging. As an example, a conversation where A sends an email to B while B replies to A by an instant message

would be represented by the instantiation of two communication descriptions, one Email Description and one IM Description. These would define roles such as addresser and addressee that are played-by instances of the class Agent representing A and B. A Communication Course will also be defined that sequences a Communication Event that has all the players of the roles defined by the particular communication description as participants. That same Communication Event will also be sequenced-by a communication course in the second (instant messaging) description having the role players within that description as further participants. Thus, conversational threads can be represented in spite of the use of different communication modes that are based on different addressing mechanisms, communication protocols, and message serializations.

4.4

Classification

The module for information classification as shown in Fig. 4 enables the reuse of classifications and concepts tackling classification issues mentioned in Sect. 2. The module is part of the core ontology for multimedia (COMM) [1] and aligned with X-COSIMO. It enables MPEG-7 compliant annotation of information as represented within X-COSIMO and explicates the annotation by semantic metadata so that it is available for any application that features information classification. A Semantic Annotation situation satisfies a Method that defines a SemanticLabelRole as well as an AnnotatedDataRole. The latter is played-by an Information Object while the semantic label role is played-by a DOLCE Particular. Accordingly, annotations of information objects with arbitrary information can be represented thus pro-

Figure 4: Classification Module viding a flexible means to incorporate different classification mechanisms ranging from user-defined, weakly formalized approaches – e.g. tagging – over semantic annotations with concepts defined in a domain ontology to annotations with instances defined within XCOSIMO itself, e.g. annotation of an email with another email or a person. For example, the concept K-Cap can play a SemanticLabelRole in semantic annotations where the AnnotatedDataRole is played-by information objects representing the K-Cap web page, the email containing the call for papers, and a research paper written for presentation at K-Cap.

4.5

Decomposition

To represent the decomposition of information, XCOSIMO is also aligned with the decomposition pattern of the COMM. The pattern can be applied to represent not only the classification of information as a whole, e.g. a file, a web page, an email, but also parts of it such as a paragraph of a web page or even a particular word in an email. The decomposition pattern can further be used to represent compound documents such as presentations that often are composed of inserted images and copied text elements. We refer to [1] for a more thorough explanation of that module.

5.

THE X-COSIM MAPPING LAYER

The reference ontology X-COSIMO is comprehensive and therefore not easy to use. Therefore, the X-COSIM architecture includes a mapping layer that encapsulates the complexity of mappings between context dependent information representations and X-COSIMO. XCOSIMA is an easy-to-use application programming interface that provides such an encapsulation. However, the X-COSIM architecture also allows for extensions of the mapping layer and does not require the use of XCOSIMA, e.g. query languages for RDF can be used to enable quick data exports and transformations from X-COSIM to context-dependent data representations.

5.1

Application Programming

To support application programming, X-COSIM provides X-COSIMA, an application programming interface for the Java language that maps (in both directions) between X-COSIMO-based and simpler, domain specific data representations. Instead of programming the comprehensive instantiations of the correct classes and properties defined by X-COSIMO and DOLCE,

developers can use simple methods provided by XCOSIMA that reflect contextual views of an application domain. For instance, the API provides a method createInstantMessage that has parameters like conversation, sender, and senderAddress. Such a method resembles methods provided by conventional APIs for dealing with message data while in fact, X-COSIMA ensures that calls of methods result in the corresponding data representation of X-COSIMO, e.g. the instantiation of a communication description that defines roles and courses which are linked to information objects and further instances of DOLCE particulars.

5.2

Data Export and Transformation

In X-COSIMO, different aspects of information are exclusively represented based on the DnS design pattern (cf. Sect. 4) that defines a formalized and general mechanism to represent context. Accordingly, a SPARQL CONSTRUCT query template can be used to transform between a data representation given by X-COSIMO to a domain specific representation conforming to some domain ontology. Listing 1 shows how to express relationships between persons as represented in X-COSIMO by means of the Friend-of-a-friend (Foaf) vocabulary2 , i.e. by using the concept of person and the knows relation defined by Foaf. Due to the standardized DnS based representation of contexts within X-COSIMO, queries for exporting to further representations follow the same pattern. The following namespace prefixes are used in Listing 1: rdf for the RDF vocabulary, dns for DOLCE, cosim for X-COSIMO, foaf for Foaf. Listing 1: Foaf Export CONSTRUCT { ? sender ? recipient ? sender } WHERE { ? desc

rdf : type rdf : type f o a f : knows

f o a f : Person . f o a f : Person . ? recipient .

dns : d e f i n e s ? a d d r e s s e r R o l e , ? addresseeRole . ? addresserRole rdf : type cosim : A d dr e ss er R ol e . ? sender dns : p l a y s ? addresserRole . ? addresseeRole rdf : type cosim : AddresseeRole . ? recipient dns : p l a y s ? addresseeRole

}

6. APPLICATION EXAMPLE We now describe a scenario that indicates how XCOSIM can be employed to implement advanced PIM support, namely we show how metadata from the communication context can be used in the context of file browsing. Several tools take part in the scenario: First, the semantics aware messenger (SAM3 ) that utilizes XCOSIMA to store metadata about incoming and outgoing messages as well as file transfers. Second, scripts that extend the Konqueror4 file manager, and third, a browsing application for RDF5 . All tools built upon X2

http://www.foaf-project.org/ http://isweb.uni-koblenz.de/Research/sam 4 http://www.konqueror.org/ 5 http://isweb.uni-koblenz.de/Research/rdfbrowser 3

Figure 7: Accessing Conversation Metadata from the File Browser

Figure 5: Instant Messaging Conversation Figure 6: Scenario Metadata (Excerpt) COSIMA for displaying, retrieving and storing metadata represented based on X-COSIMO. The scenario: Consider two persons A and B that have an instant messaging conversation. Within the conversation, A asks B about interesting recent research papers and B sends a paper to A (cf. Fig. 5). In a further conversation, A sends the received paper as email attachment to another person C. The information lifecycle of the scenario is described as follows: For each instant message/file sent and received by A, SAM calls methods provided by X-COSIMA to create communication metadata, e.g. createInstantMessage. The calls result in a new Communication Situation that satisfies an Instant Messaging Description with all its roles and courses (cf. Fig. 6). RDF resources representing A and B are created (if not yet existing) and are related to addresser and addressee roles, new Information Objects are created that play the roles of a transferred file, message, or address. New Digital Realizations are created that describe the storage of the transferred file in a local folder, or the realization of an address as an RDF literal. When A sends an email to C with the file as attachment, the email client of A creates a new email description using X-COSIMA to create ana-

logue instantiations for the email-based conversation.6 Fig. 6 exemplifies the created metadata and indicates the boundaries of the different descriptions that reify instances of DOLCE concepts that take part in the scenario. As shown in Fig. 7, when A now browses the file system with Konqueror and performs a right-click on the file xcosim.pdf, a context menu is shown that contains entries for viewing X-COSIM based metadata about conversations, classifications, and realizations associated to the selected file. For instance, A can click the conversations menu entry to see that the file x-cosim.pdf has been sent by B within an instant message conversation and that it was sent to C by email (cf. Fig. 7).

7.

RELATED WORK

Analysing solutions for complex task support provided by existing research prototypes of semantic desktops is difficult. Few documentation about strategies and solutions employed by existing systems is available and 6

A plugin for an email client is currently being developed. Its functionality will be analogue to that provided by the SAM plugin

current systems often focus on technological and user interface aspects. Accordingly, our review of related work is based on a few publications and the analysis of non-scientific resources such as development guides and howtos. None of the systems we reviewed provides support for information reuse across contexts. They make use of the resource description framework7 to relate information to each other and employ schema and ontology languages to model information. However, they do not answer the question of how to consistently represent and map PIM relevant information so that it can be interlinked and reused in complex personal work processes. Similar to the X-COSIM approach, the Gnowsis [11] and Iris [2] semantic desktops are based on the use of a reference ontology intended to provide a consistent representation of different information relevant in PIM. The employed ontologies, however, lack the thorough analysis given by foundational ontologies and thus lack the broad scopus that is required to establish information linkage and reuse across work contexts. Both Gnowsis and Iris do not provide mechanisms for capturing different, potentially conflicting aspects of the same information that need to be captured in order to map from the reference ontology to context dependent conceptualizations employed by individual PIM applications. With Haystack [7], different pieces of information can be organized (e.g. in collections) and associated to each other. They can be reused in particular Haystack modules. Plugins, however, implement their own, application specific models that do not necessarily conform with the conceptualization of existing and future Haystack plugins. Accordingly, reuse of information provided by one plugin in another plugin requires a oneto-one mapping between the data models employed by each plugin. This is not realistic, as one may not assume that one plugin manages all the metadata provided by all the other PIM applications. There currently exist several ontologies to represent particular portions of information relevant for PIM, modeling contact and address data or social connectivity. The Foaf vocabulary, the ontology for vCards8 , and the W3C PIM vocabulary9 are examples of those. Such domain ontologies are not suitable to provide a reference model for arbitrary personal information for which they lack the broad scopus of foundational ontologies. However, they express contextualized information models that determine the design of future extensions of X-COSIMA.

8.

CONCLUSION

In this paper we have presented X-COSIM, a framework for cross context semantic information manage7

http://www.w3.org/RDF/ http://www.w3.org/2006/vcard/ns 9 http://www.w3.org/2000/10/swap/pim/contact 8

ment. The ontology X-COSIMO, together with its basis DOLCE, and the provided API X-COSIMA constitute a methodological basis for developing ontologies and applications for a really seamless semantic desktop. They provide a sound conceptual foundation that may span the whole lifecycle of complex tasks, but by their modular architecture they remain extensible towards new PIM applications. Among the next challenges in semantic desktop research we see the development of methods and tools that exploit metadata as provided by X-COSIM to establish new desktop services that go beyond the mere presentation of context information. As a further challenge, we consider the extension of the semantic desktop towards the web, raising privacy and security issues related to metadata exchange and propagation across desktops.

9.

ACKNOWLEDGMENTS

This work was funded by the X-Media project (www.xmedia-project.org) sponsored by the European Commission as part of the Information Society Technologies (IST) programme under EC grant number IST-FP6026978.

10.

REFERENCES

[1] R. Arndt, R. Troncy, S. Staab, L. Hardman, and M. Vacura. Comm: Designing a well-founded multimedia ontology for the web. In ISWC, 2007. [2] A. Cheyer, J. Park, and R. Giuli. IRIS: Integrate. Relate. Infer. Share. In Workshop On The Semantic Desktop, ISWC, 2005. [3] S. Decker. The social semantic desktop: Next generation collaboration infrastructure. Information Services and Use, 26(2):139 – 144, 2006. [4] A. Gangemi and P. Mika. Understanding the semantic web through descriptions and situations. In CoopIS/DOA/ODBASE, pages 689–706, 2003. [5] S. C. Herring. Computer-mediated communication on the internet. Annual review of information science and technology, 36:109–168, 2002. [6] R. Jakobson. Style in Language, chapter Closing statement: Linguistics and Poetics, pages 350– 377. Cambridge, Mass., MIT Press, 1960. [7] D. R. Karger, K. Bakshi, D. Huynh, D. Quan, and V. Sinha. Haystack: A General-Purpose Information Management Tool For End Users Based On Semistructured Data. In CIDR, pages 13–26, 2005. [8] S. L. Kogan and M. J. Muller. Ethnographic study of collaborative knowledge work. IBM Systems Journal, 45(4):759, 2006. [9] W. E. Mackay. More Than Just a Communication System: Diversity in the Use of Electronic Mail. In CSCW, pages 344–353, 1988. [10] C. Masolo, S. Borgo, A. Gangemi, N. Guarino, and A. Oltramari. Wonderweb deliverable d18 ontology library (final). Technical report, 2003. [11] L. Sauermann, G. A. Grimnes, M. Kiesel, C. Fluit, H. Maus, D. Heim, D. Nadeem, B. Horak, and A. Dengel. Semantic desktop 2.0: The gnowsis experience. In ISWC, pages 887–900, 2006. [12] S. Whittaker, V. Bellotti, and J. Gwizdka. Email in personal information management. Commun. ACM, 49(1):68–73, 2006.