User Interfaces for Semantic Web Applications

Abstract [top]

This paper discusses the role of user interaction in various semantic Web applications, mainly focusing on two major application use domains: information retrieval and information management. User interfaces of two existing applications (Haystack and mSpace) from each of these use domains have been discussed in order to spotlight various issues in user interface design for semantic Web applications. These issues can serve as guidelines for future interface design for applications using semantic Web technologies.

Keywords [top]

User interface design, semantic Web, semantic information, semantic Web applications, information management, information retrieval, knowledge management

Introduction [top]

Growing information on the Web has made it increasingly difficult to find, access, present, and maintain different types of information. The main reason behind this was identified as presentation of information content in natural language understandable only to humans, who are comparatively less efficient than machines (or agents) in performing above mentioned tasks [Fensel et al., 2002]. The semantic Web project [Semantic Web project] is among many other research projects undertaken to deal with this problem. Tim Berners-Lee, the creator of the World Wide Web (www), has envisioned semantic Web as an extension of the current Web, adding structure and semantics to it, so that Web information can also be processed by computer agents [Berners-Lee et al., 2001]. This structure will be added by expressing relationships among different information entities on the Web. These relationships can be articulated in a navigational space with each resource or information object described through a unique URI (uniform resource locator) and a mapping from URI to its resources. Anything can be identified as a resource in this case; ranging from content of a Web page to a handheld computer or a real person.

Underlying goal of the semantic Web project is to make the Web more usable for its users to facilitate their primary activities of information retrieval and information management. User interface, through which users gain access to semantic information, is a key in achieving this goal. In order to take advantage of the added structure and semantics to the Web content, we need extended user interaction experience through innovative user interfaces [Huynh et al., 2003]. Various tools and applications have been developed to assist users create, manipulate, retrieve, present, organize, and manage semantic information. Semantic Web tools are mainly designed to aid content creators create semantic information through different design activities like designing or visualizing ontologies [Protégé, Ontomat-Annotizer, and Smile project], creating RDF [RDF primer] files for resources, searching semantic data [Swoogle], and creating semantic meta data. Semantic Web browsers let users collect and browse semantic information linked from ordinary Web pages [Piggy-Bank]. On the other hand semantic Web applications are more focused towards various information activities of end users [Haystack project, mSpace project, Gnowsis, Fenfire, Flamenco search interface project, Topia, and AKT CS AKTiveSpace]. These applications can be categorized into four major use domains: information retrieval, information organization, information presentation, and information management including personal knowledge management. Among these domain areas informati on or knowledge management has been identified as one of the most influenced area enabled by semantic Web technologies. Traditional knowledge management is now facing problems in the current Web due to information overload, inefficient keyword searching, heterogeneous information integration, and location independent access [Fensel et al., 2002a], which have been tackled by semantic Web technology [Fensel, 2001 and Fensel et al., 2002b]. Other domain areas tend to overlap each other in different applications to satisfy extended user needs.

In this paper, I discuss various issues in user interface design for semantic Web applications, primarily in information management and information retrieval domains, by examining existing user interfaces in this arena. The next two section of the paper discusses functionalities and interface design of two major semantic Web applications and get to various issues that can facilitate future semantic Web interface design. Terms knowledge and information have been used interchangeably in this paper so as to broaden their scope of use in various applications.

Existing Semantic Web Application User Interfaces [top]

The design of the user interfaces for Web applications has always been of great importance. The main reason behind this has been that the user frequently does not receive training in the use of these applications as in the case of most traditional software systems. Therefore, a web interface has to be intuitive as the user must be able to self-train him on the application. Now with the advent of semantic Web technologies design of intuitive interfaces has becoming more and more significant as users will face interfaces that they haven not seen or used before. Semantic Web researchers have been working on these innovative interfaces to make semantic Web applications more usable.

In this section I will discuss semantic Web applications focusing on the user interface design. Applications discussed here are primarily targeted towards two major domains of semantic Web application use: information management and information retrieval. Semantic Web applications in information management area allow users to browse, organize, and manage different types of information (like e-mail messages, instant messages, contact information, web pages, documents, news, bibliographies, annotations, music, images, blog feeds, audio/video data, pictures, etc.) from a single interface. These interfaces remove the barrier laid by different dedicated applications and allow a single point of access to different types of information. They identify different types of information as a unique information object and present relationships between them for effective information organization and management. These relationships are described in terms of ontologies by translating original data into RDF format. Haystack (discussed below) and gnowsis [Gnowsis] are some of the applications that fall into this category. Further, semantic Web applications in information retrieval use domain allows users to effectively search through different types of data by articulating various aspects of the same data and providing relationships between them. Interfaces in this domain categorize bulk of information, usually of the same data type and generate previews into a single window, where users can navigate through what they are looking for. It also allows personalized organization of this information from the same interface. These applications are mainly concentrated in a particular type of information like images, audio, video, but can also be extended for variety of types. mSpace (discussed below) Flamenco [Flamenco search interface project], Topia [Topia], and AKTiveSpace [AKT CS AKTiveSpace] are some of the applications that fall into this category.

mSpace

mSpace [mSpace project and Harris et al., 2004] is a semantic Web interface developed at School of Electronics and Computer Science (ECS) at the University of Southampton to facilitate information access, browsing, and organization given that the user has limited domain knowledge. This is achieved through exploring various relationships in information through semantic Web technologies, and allowing users to manipulate information categorization to suit their interests. Researchers developed a demo version (mSpace classical music browser) of mSpace framework to access and browse music information, but the framework can also be applied to any type of information.

mSpace interface design is oriented towards three main functionalities [see figure 1]. First, the interface allows loads of information categorized into a single window, where users can browse through what they are looking for. In the music browser information is categorized by era, composer with composer country and death date, music arrangement, and the music piece. Users can browse through each of these categories and also access associated information to develop domain knowledge; all in one window. The interface also provides good starting points to help users in their search process. Second, the interface provides information specific preview cues while browsing. In the music browser users can actually listen to the piece from the same place to create a searching strategy. Third, it allows personalized organization of information. In the music browser interface, users can swap, slice, sort, add, and remove categories the way they want from the menu triggered by right click. Thus by providing domain knowledge and personalized organization of information, mSpace allows users to build knowledge through exploring new relationships between data which can enhance their information search process. mSpace framework uses semantic Web technologies behind this interface to represent complex information space with categorization, previews, organization and relationships in data.

Figure: 1 - mSpace interface

Haystack

The Haystack project [Haystack project] is an undergoing research application of knowledge access user technologies component of the research project Oxygen [Project Oxygen], at MIT Laboratory for Computer Science and Artificial Intelligence (CSAI). Haystack seeks to apply semantic web technologies to personal information management. It allows individuals to manage their information in the ways that makes the most sense to them.

Haystack interface [see figure 2] exhibits three novel functionalities to facilitate users in their information management tasks. First, it incorporates and exposes all types of information in a single, coherent manner. It provides a single, uniform interface for viewing and organizing of e-mail, instant messages, contact information, web pages, documents, news, music, images, blog feeds, etc. Second, in haystack every entity whether it is a simple text in an email message or an email message itself is considered as an information object. Any of these information objects can be right clicked for its context menu, allowing immediate access to all the operations that make sense for that object. The interface also lets users define their own information objects to incorporate any non standard types of information. Users can readily define attributes of these new objects that help them categorize and retrieve information, and add new relationships to objects. An information object or operation can also be downloaded from outside applications which will be immediately available to use. Finally, haystack gives users flexibility to modify standard as well as user defined information objects irrespective of its type and application it belongs to. For example an operation (such as spellchecking, sending an e-mail message, or rotating an image) can be invoked at any time on any object for which the operation makes sense to the user, not limited to the specific application (such as text editor, e-mail client, or graphic application).

Figure: 2 - Haystack interface

Potential Issues in User Interface Design for Semantic Web Applications [top]

Following user interface design issues are identified from the above discussion on user interfaces for existing semantic Web applications.

"All in one" versus dedicated interfaces

One of the central themes of the semantic Web interfaces discussed above was providing a single interface to access, organize, search, and manage different types of information. These interfaces tend to break the barrier laid by dedicated applications to access different types of information and perform various application specific functions. For example in traditional applications, email messages can only be viewed or edited through an email client, instant messages with instant messaging client, and Web browsing and searching was efficiently done only in the Web browser. Further, non internet application functions like managing contacts and schedules, organizing documents on the hard disk were only done through their dedicated applications. Semantic Web interfaces provide a unified interface to perform these dedicated application specific functions from a single point. As an interface designer for semantic Web applications we need to consider several questions as they relate to "all in one" interfaces. Is it always good to have "all in one" interface for all the functions that can be done individually? What if users are looking for in depth functionalities for a particular application, which these "all in one" interfaces are not capable of providing? We also need to decide which all functionalities should be integrated into "all in one" interfaces and which not. What should be the driving functionality of such interfaces [Economist, 2005]? For example whether these interfaces should primarily focus on emails and Web browsing and provide additional functionalities in accordance with them or focus on scheduling and contact management and provide Web browsing and emails as secondary functions.

The Author in [Economist, 2005] talks about various such issues with respect to "all in one" electronic devices versus dedicated devices. The author makes a point that usually convergence of different functionalities into one device leads to surprising-looking offspring device with all the different functionalities but it makes the device more complex and hard to use and sometimes increases the cost also. Same concept can also be applied to user interfaces. "All in one" user interfaces can mislead users in different functionalities and make the interface more difficult to use.

Which application use domain to focus?

Four semantic Web application use domains: information retrieval, information organization, information presentation, and information management were identified earlier in the paper. While designing semantic Web application interfaces one of the issues to consider is which domain of use to focus in the application. Discussion on existing semantic Web applications individually focused on information retrieval and information management. So the question to consider here is should we concentrate on a single domain for different applications or a single application should have functionalities supporting multiple domains?

Integration with existing semantic information

As noted earlier, Semantic Web is an extension to the existing Web where structure and semantics will be added to the current Web content [Berners-Lee et al., 2001]. It is important to keep in mind that the newly created interfaces will also be used with non semantic information already present in the current Web. Thus these interfaces should also support older content by providing efficient integration of semantic and non semantic information.

Support for all the resource types

Every resource whether it is a Web page or a physical device, a unique identifier is assigned to all of them in order to identify them in the semantic Web navigational space. Semantic Web interfaces should allow access and manipulation of all the resources in addition to the Web content. The information object concept in the Haystack interface works kind of similar to this. Haystack identifies every information type and operation as an information object. This information object can be anything ranging from a Web page, email message, or a personal contact. Various operations that users perform in an application like sending an email, editing an image etc. are also considered as information objects and can be handled similarly. This concept of information object can be extended to include other types of resources, like physical devices as PDAs, cell phones, etc. or every entity on the Web like an item sold on Amazon.com, in the same interface supporting all the possible interface functionalities.

Providing relationships between different resources

The main goal of semantic Web is to provide relationships between various resources and identify them in a common space. In semantic Web applications it is essential how these relationships are visualized, managed, organized, and accessed in a particular application use domain. For example in information retrieval domain, mSpace interface presents information in different categories and offers various preview cues to search efficiently in a large domain specific information space. This categorization and previews are mainly based on various relationships identified and organized trough RDF, among variety of information, from a single window. This helps users in determining better starting points for searching.

Personalization in providing semantic context

Interfaces that support user centered customization can be more effective in serving them than user interfaces which are designed for general user community [Karger and Quan, 2004]. Haystack allows users to customize standard as well as user defined information objects irrespective of its type and application it belongs to. Similarly mSpace lets users personalize information organization by modifying various categories the way they want to satisfy their information needs. Thus in designing interfaces for semantic Web applications users should be given some control to set preferences that suits their needs. In addition to presenting information in customized views, interfaces should also be flexible enough to allow users to modify data and various operations on data. Further, interfaces should also be adaptable to understand user's information behavior throughout his or her interaction with the application to make the interface more usable down the line.

One of the most important components of the semantic Web is computer agents. They work in coordination with humans to make sense of semantic information on the Web [Berners-Lee et al., 2001]. Lieberman [Lieberman, 1997] identifies various design principles for autonomous interface agents. In addition, Norman [Norman, 2002] presents various issues on how might people interact with agents so their interaction with users can be designed effectively.

Conclusion [top]

Semantic Web technologies add structure and semantics to the existing Web by presenting relationships between different resources on the Web. Semantic Web applications provide access to large amount of information in a variety of different types; therefore effective information retrieval, organization, and management are considered key factors in interface design for these applications. This paper identifies potential issues in user interface design for semantic Web applications mainly in information retrieval and information management domains. These issues can serve as design guidelines for future interface design for semantic Web applications in these two domains.

References [top]

AKT CS AKTiveSpace]: AKT CS AKTiveSpace. Advanced Knowledge Technologies. http://triplestore.aktors.org/SemanticWebChallenge/
Berners-Lee et al., 2001]: Berners-Lee, T., Hendler, J. and Lassila, O. (2001). The semantic Web. Scientific American, May. Retrieved April 19, 2005, from http://www.scientificamerican.com/print_version.cfm?articleID=00048144-10D2- 1C70-84A9809EC588EF21
Economist, 2005]: The device that ate everything? The Economist, March 10, 2005. Retrieved April 19, 2005, from http://economist.com/science/tq/displayStory.cfm?story_id=3713943
Fenfire]: Fenfire. http://fenfire.org/
Fensel, 2001]: Fensel, D. (2001). Ontologies: silver bullet for knowledge management and electronic commerce . Springer-Verlag, Berlin.
Fensel et al., 2002a]: Fensel, D., Bussler, C., Ding, Y., Kartseva, V., Klein, M., Korotkiy, M., Omelayenko, B., and Siebes, R. (2002). Semantic Web application areas. Proceedings of 7th International Workshop on Applications of Natural Language to Information systems (NLDB 2002) .
[Fensel et al., 2002b]: Fensel, D., Hendler, J., Liberman, H. and Wahlster, W. (eds.) (2002). Semantic Web technology . MIT Press, Boston, MA.
[Flamenco search interface project]: The Flamenco Search Interface Project. School of Information Management and Systems (SIMS) at the University of California at Berkeley. http://bailando.sims.berkeley.edu/flamenco.html
[Gnowsis]: Gnowsis. http://www.gnowsis.org/
[Harris et al., 2004]: Harris, C., Owens, A., Russell, A. and Smith, D. A. (2004). mSpace: Exploring the semantic Web. Unpublished project report. University of Southampton.
[Haystack project]: The Haystack Project. MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). http://haystack.csail.mit.edu
[Huynh et al., 2003]: Huynh, D., Quan, D. and Karger D. R. (2003). User interaction experience for semantic Web information in The Twelth International World Wide Web Conerence , May 20-24, Budapest, Hungary.
[Karger and Quan, 2004]: Karger, D. R. and Quan, D. (2004). Prerequisites for a personalizable user interface in Intelligent User Interfaces (IUI) 2004 Workshop on Behavior-based User Interface Customization , January 13-16, Madeira, Portugal.
[Lieberman, 1997]: Lieberman, H. (1997). Autonomous Interface Agents. ACM Conference on Human-Computer Interface [CHI-97] , Atlanta, ACM Press.
[mSpace project]: The mSpace Project. School of Electronics and Computer Science (ECS) at the University of Southampton. http://mspace.fm/
[Norman, 2002]: Norman, D. (2002). How Might People Interact with Agents? Retrieved April 19, 2005, from http://www.jnd.org/dn.mss/agents.html
[Ontomat-Annotizer]: Ontomat-Annotizer. http://annotation.semanticweb.org/ontomat/index.html
[Piggy-bank]: Piggy-bank Mozilla Firefox extension, The Smile Project. W3C, HP, MIT Libraries, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). http://simile.mit.edu/piggy-bank/index.html
[Project Oxygen]: Project Oxygen. MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). http://www.oxygen.lcs.mit.edu/index.html
[Protégé]: Protégé. Stanford University. http://protege.stanford.edu/
[RDF primer]: RDF Primer. World Wide Web Consortium. http://www.w3.org/TR/REC- rdf-syntax/
[Semantic Web project]: The Semantic Web Project. World Wide Web Consortium. http://www.w3.org/2001/sw/
[Smile project]: The Smile Project. W3C, HP, MIT Libraries, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). http://simile.mit.edu/
[Swoogle]: Swoogle. http://www.swoogle.com
[Topia]: Topia. http://topia.demo.telin.nl/