An Architecture For Character-Mediated Interactive Presentations
0 likes7 views
This document describes the architecture of Cyrano, a system for interactive presentations of cultural heritage information using an artificial character. Cyrano generates presentations on a portable device where a character inhabits a virtual space and responds to user inputs with pre-built script units. The system architecture is based on beliefs, desires, and intentions (BDI) and uses modules to understand user input, determine goals, select appropriate scripts, and schedule multimedia output. An example application for the Royal Apartments of La Mandria is provided to illustrate how the system works in a drama-based interactive presentation.
An Architecture For Character-Mediated Interactive Presentations
- 1. An Architecture for Character-Mediated Interactive Presentations Rossana Damiano1 , Francesca Biral2 , Vincenzo Lombardo1 , Antonio Pizzo2 1 Dipartimento di Informatica and Centro di Scienza Cognitiva, Università di Torino c.so Svizzera 185, Torino, Italy {vincenzo, rossana}@di.unito.it 2 Dipartimento di Discipline Artistiche, Musicali e dello Spettacolo, Università di Torino via S. Ottavio 9, Torino, Italy {antonio.pizzo, francesca.biral}@unito.it Abstract. In this paper, we present the architecture of Cyrano, a system for character-based, drama-like interactive information presentation in the specific domain of cultural heritage. The theoretical approach, which follows the interactive drama paradigm, responds to a well-documented need for a drama-based metaphor to deal with cultural heritage presentation issues. Cyrano is implemented on a PDA-wireless network infrastructure adapted to the constraints of an historical site environment, and takes into account visitors’ requirements, as described in literature on PDA applications in museums and historical sites. 1. Introduction In recent years, much attention has been devoted to artificial characters as a valuable communication metaphor in information presentation (see papers in Cassel et al. 2000): character-mediated presentations are considered to be more compelling, especially if they are delivered in the form of drama and entertainment (Gershon & Page 2001). The dramatic component is fundamental, since, as Picard points out (Picard 1997), an appropriate and well-calibrated level of emotional arousal helps focalizing the attention and facilitates the understanding. In this paper, we present the architecture of Cyrano, a system for character-based information presentation based on an interactive drama paradigm (Kelso et al. 1992, Mateas 2001& 2002). Interactive drama relies on believable agents, i.e. “broad” and shallow agents which display a range of abilities (including reactivity, emotions, social and language skills) that convey the user the illusion of reality and allow for her/his suspension of disbelief. This paradigm, created for entertainment purposes, provides a valuable tool in the application design of electronic guidebooks, an area where research on interaction metaphors is so far lacking. Cyrano generates interactive presentations acted out by an artificial character who inhabits a virtual space situated on a portable device. The system architecture, typical of BDI agents, exploits a repository of pre-built script-units, that can be combined in order to react efficiently to the user’s inputs, mainly speech (natural language) and hardware buttons pressing. The motivation for having pre-built script-units is that the simulation of multimodal systems and their synchronization is a very costly task, and their on-line generation may be ineffective at the current state of the art, i.e. characters may appear un-reactive if not “dumb” (Cassel et al. 2000), therefore conflicting with the interactive drama paradigm. The paper is organized as follows: in the next section after a brief survey of electronic guidebook applications we point out the lack of a communication metaphor in the application design; then we illustrate the Cyrano architecture; finally, we show a quick overview of the DRAMA-Tour application, developed in the context of an historical site. 2. The context: a brief survey of PDA applications in museums and historical sites Applications featuring hand-helds in a museum context date as far back as 1993, the year the first P.D.A. (Personal Digital Assistant), the Apple Newton, was launched on the market. Major drawbacks reportedly cursed early prototypes: the Newts were heavy, fragile, expensive and museum staff was even reluctant to give them out fearing “equipment malfunction and theft” (Amirian 2001). The same range of problems, though, can be encountered in reviewing even recent applications: PDAs are still quite expensive, fragile and their use in a museum context can prove a real challenge for designers and visitors (Electronic Guidebook Forum, 2001). Nevertheless, the PDA has been labelled the next killer app in museums (Schwarzer 2001). Research on the subject is thriving. Computer scientists and the IT community in general regard the museum visitor as the typical specimen of mobile computing user. Since research on the subject mainly streams from the concern of testing emerging technologies such as wireless networks and mobile computing devices, the museum is often regarded as the mere environment where to test a one-off, short-lived prototype. However, museums strive to attract a larger, and more socially diverse, public, as their role in modern society is shifted from being a repository to
- 2. being communicator of culture. The most enlightened museum directors and curators are aware of the challenges this role shift involves and look upon technology as a useful tool to implement effective communication strategies. Spalding has vividly described a futuristic scenario where museums take advantage of mobile computing technologies in order to efficiently perform story-telling on the invaluable and fascinating objects that they own (Spalding 2002). The technological possibilities to make Spalding’s scenario come true already exist (see Biral & Lombardo 2003) for a review. However, there is a need to further investigate the application design in the terms of constructing an effective metaphor for the visit. Most applications do not feature a general metaphor for the visit and rely on a browser metaphor for the interface design, a choice liable of generating confusion for the user, e.g. reportedly, users who followed links to exhibitions they were not at experienced a “lost in hyper-reality” phenomenon (Fleck et al. 2002). Xerox PARC researchers can be said to be the only team who has come up with an effective metaphor for hand- helds use in a museum context (historical site). They adopted an innovative perspective by giving the electronic guidebook a conversationalist role, in order to enhance socialisation between visiting companions, and developed a prototype that permitted eavesdropping on the audio content of a companion's guidebook. The conversationalist metaphor (and eavesdropping) proved effective for “building stronger interactional ties between companions (encouraging more natural conversations) as well as increasing awareness of the room and its contents” (Grinter et al. 2002). Moving on from this successful conversationalist metaphor we are developing an application for an historical site visit where the ‘conversationalist’ is embodied in an artificial character who, according to loose dramatic constraints, performs the role of a former inhabitant of the site. A drama-based metaphor is able to address critical problems in the often “lifeless” and “squalid” presentation of cultural heritage sites (Waterfield, 1998). It is well-known that theatre is an ideal metaphor to enhance visits in an historical site setting, as it is currently used in exhibition design at various levels, because of the intrinsic, mute nature of objects (Maure, 1995). Moreover, costume drama and ‘ghost’ performances, massively employed in historical sites, heritage centres, period rooms and the like to “bring the past back to life” are hugely popular amongst the public, although at times criticized by historians on account of being attempts in re-writing history more than re-enacting it (Buckley 1996). 3. System architecture The system is intention-driven, as the interactional and dramatic goals of the system are explicitly represented, and drive the behaviour generation process. The system works by selecting a script according to its presentation goals (Cohen & Levesque 1990, Bratman et al. 1988) and monitoring its execution so as to react to user input (Wooldridge & Parsons 1999). A script consists of a sequence of pre-recorded audiovisual units (script-units), a set of pre-conditions, that determine the applicability of the script, and a set of effects, that hold after the script execution. Scripts are stored in a script library, the Script-KB (see Figure 1). The system is reactive: if the current script - previously selected as contextually appropriate to the system goals - is not appropriate anymore, the system switches to a new script. However, the system can react only within a set of well-identified conditions, in order to preserve the dramatic coherence encoded in scripts by the application designer. During the deliberation phase, the system will abandon the execution of the current script only if a set of “interrupt” conditions hold: if the user has abruptly changed location or has shifted attention through the appropriate input devices. Now we first describe the basic structure of script-units and then the Cyrano architecture. Proceeding bottom-up, we define Clip any multimedia item, containing audio-video information, graphics or animation. A Script-unit is an elementary behavioral unit featuring a single character (Character clip), plus any additional audio-visual information (e.g. Context-clips provide extra information, like background or sound effects). Script units are concatenated to form sequences (Scripts in the system terminology) by interleaving Script-units and Transitions-units. The role of Transition-units is twofold: they produce the system behavior while waiting for the user input, and guarantee a smooth transition between subsequent Character-clips. In this way, the developer is not constrained to worry about graphical coherence when designing Character-clips, and Character-clips can be sequenced to form different Scripts with some degree of freedom. The system architecture is structured as a cascade of modules (see Figure 1). Each module takes into account the output representation of the previous module and the interaction already occurred in the current session (interaction history - IH), and produces a new representation while upgrading the interaction history. The Input Manager takes as input various signals from the input devices (user requests expressed via text/speech, tapping, buttons, and the information about the device current location), and converts it into a logic representation. Based on this representation, the Understanding Module identifies the user's goals: for example, the user may desire to obtain more information about a certain item (information seeking goals), ask for clarification about a system turn (interaction handling goals), or simply signals the willing to attend the remaining presentation. Then, the Goal Manager generates the system goals in response to the identified user’s
- 3. goals by consulting the Goal-KB, a set of rules which select the interactional and dramatic goals of the system based on the interaction context and on the advancement of the drama. The Script Manager, the deliberative core of the system, exploits the knowledge encoded in the Script-KB, and the input from the previous modules (Interaction History, User’s and System’s goals) to deliberate about the currently selected script, by continuing its execution with the next script-unit, switching to a new script, or retrieving a script (if no script has been previously selected). If a new script has to be selected, the Script Manager consults the Script-KB in search for a new script with applicable preconditions. Finally, the Presentation Module identifies a suitable transition (transition-unit) from the previous script-unit to the current one. The Action Scheduler receives a pair <transition-unit, script-unit> from the Presentation Module and dispatches it to the multimedia output devices for synchronised execution. Figure 1: Architecture of the Cyrano system 4. An application in an historical site context We are currently developing a prototype application for the visit of the Appartamenti Reali de La Mandria (Royal Apartments of La Mandria) in Venaria, Turin, Italy. It is a bourgeois 14-room apartment where Vittorio Emanuele II, the first of re-united Italy’s kings, spent 30 years. The application infrastructure is made of PDAs (equipped with head-phones and micro-phones) and a wi-fi LAN, with all the pre-packed clips residing on the PDAs and all the architectural components hosted on the site’s server. We have relinquished plans to tag Input devices: LOCATION PEN POINTING SPEECH BUTTONS Input Manager Understanding Module Goal Manager Script Manager Action Scheduler Output devices: CHAR ANIMATION GRAPHICS SPEECH TEXT Hardwired reactions Presentation Module GOAL- KB System Goal SCRIPT-KB User Goal Script-unit Input Representation Transition-unit + Script-unit Units Execution Interaction History TRANSITIONS POOL
- 4. individual exhibits (a solution that is seldom feasible in historical sites) and settled on a more coarse-grained localisation mechanism that uses strength of wi-fi signal (closeness to access points) to triangulate the visitor’s location. On perceiving the visitor’s location the system executes one of the Scripts relative to the Goal of presenting that particular location. We provide here a brief example of a drama-based presentation of a room in the king’s residence, the Billiards Room. In the system execution the current location of the user is the Billiards Room; the current system goal is the presentation of the Billiards Room where this presentation can be conveyed throughout several Scripts. We describe the Script-units of a specific Script, called “PlayBilliards”, where the king character Vittorio Emanuele II exhibits a playful and competitive attitude towards the user. Let L.S., M.S. and C.U. stand for long shot, medium shot and close-up, respectively. The description of the first Script-unit and the following Transition-unit are in figures 2 and 3 (see the pictures from our Macromedia Flash implementation). SCRIPT UNIT 1 CONTEXT CLIP CHARACTER CLIP THE KING ENTERS FROM RIGHT AND STOPS IN THE MIDDLE OF THE STAGE START M.S. OF THE KING WITH BILLIARDS CUE IN HIS HANDS THE KING DO YOU PLAY BILLIARDS? BODY A WALZER MUSIC. BACKGROUND: BILLIARD ROOM TAPESTRY THE KING RAISES HIS EYEBROW QUIZZICALLY (M.S.) END Figure 2 –Script-unit 1 and three frames from the Character clip. TRANSITION UNIT 1: LOOPABLE CLIP “CHALK” CONTEXT CLIP CHARACTER CLIP THE KING HAS HIS EYEBROW RAISED QUIZZICALLY (C.U.) START THE KING PUTS CHALK ON THE TOP OF HIS BILLIARD CUE BODY BACKGROUND: BILLIARD ROOM’S TAPESTRY “EMPTY CLIP” END Figure 3 – Start-clip and Body-clip for Transition-unit 1 The Body-clip of Transition-unit 1 is a Loopable-clip, that takes into account the user’s interaction time (in this case the system awaits for a positive or negative answer) and the eventual inertia of the system in calculating the best dramatically suited Script-unit to follow in response of the user’s selection.
- 5. 5. Conclusions The functioning of the system is modelled on the the way script writers work. The task of writing a drama can be divided into two main steps: the author first creates the characters by implicitly defining their behavior in a given set of situations, then sets up a sequence of events to which they react, thus obtaining the drama. Similarly, in our system, the developer of a drama application first encodes the system's possible behaviors into a set of scripts, leaving to the system the task of generating an intentional behavior by selecting a script among the available ones. From the viewpoint of the development of applications, this approach presents some advantages which derive from its resemblance to the human authoring process. At the script-unit level, encoding the system's behavior into pre-built, elementary units, enforces the tendency of script writers to reason in terms of complex behavioral units, and, at the same time, it alleviates the difficulty of generating basic behaviors through a number of coordinated cognitive system. At the script level, the necessity to organize the behavioral units into a coherent sequence, allows the developer to follow those well-established narrative and interactional macro-structures which drive both the authoring and understanding processes. The obvious drawback of this approach is that the system, due to the prominent role of off-line behavior definition, manifests a certain amount of rigidity with respect to the agent-based systems, like Oz (Mateas 2001, 2002). Bibliografy (Amirian 2001) S. Amirian, Hand-held Mobile Computing in Museums. A Background Paper, 2001 Web link: http://www.cimi.org/whitesite/AmirianBJM.htm. (Bratman 1988) Bratman, M.E., Israel, D.J., and Pollack, M.~E. Plans and resource-bounded practical reasoning, Computational Intelligence, 4:349-355, 1988 (Buckley 1996) A. Buckley, in G. Kavanagh (Editor), Making Histories in Museums, Leicester University Press, 1996. (Cassell, 2000) J. Cassell, J. Sullivan and S. Prevost and E. Churchill, Embodied Conversational Agents, The MIT Press, Cambridge, Massachussets, 2000 (Cohen & Levesque 1990) P. R. Cohen and H. J. Levesque, Intention is choice with commitment. Artificial Intelligence, 42:213-261, 1990. (Gershon & Page 2001) N. Gershon and W. Page, What Storytelling Can Do for Information Visualization, Communications of the ACM, p. 31-37, vol. 44(8), 2001. (Grinter et al. 2002) R. E. Grinter, P. M. Aoki, A. Hurst, M. H. Szymanski, J. D. Thornton and A. Woodruff, Revisiting the Visit: Understanding How Technology Can Shape the Museum Visit, in Proc. ACM Conf. on Computer Supported Cooperative Work, New Orleans, LA, Nov. 2002, 146-155. (Kelso et al. 1992) M. Kelso, P. Weyhrauch, J. Bates, Dramatic Presence, 1992. (Fleck et al. 2002) M. Fleck, M. Frid, T. Kindberg, R. Rajani, E. O'Brien-Strain, M. Spasojevic, From Informing to Remembering: Deploying a Ubiquitous System in an Interactive Science Museum, Technical Report HPL2002-54, HP Labs Palo Alto, 2002. (Mateas 2001) M. Mateas, M. Leonardo, Expressive AI: A hybrid art and science practice, Journal of the International Society for Arts, Sciences, and Technology, 34 (2), 2001. 147-153. (Mateas 2002) M. Mateas, Ph.D. Thesis, Interactive Drama, Art, and Artificial Intelligence, 2002 (Maure 1995) M. Maure, The Exhibition as Theatre, On the Staging of Museum Objects, Nordinska Museologi, 1995, pag. 4. (Picard 2000) R. Picard, Affective Computing, The MIT Press, 1997. (Schwarzer 2001) M. Schwarzer, Art & Gadgetry. The Future of the Museum Visit, Museum News, July/August 2001. (Spalding 2002) J. Spalding, The Poetic Museum. Reviving Historic Collections, Prestel 2002. (Wooldridge & Parsons 1999) M. Wooldridge and S. Parsons, Intention reconsideration reconsidered. In Mueller, J., Singh, M.P., and Rao, A.S., editors, Proc. of ATAL-98, volume 1555, pages 63--80. Springer- Verlag, 1999.