A pre-condition for the construction of guided tours based on user tracking is that navigation consists of a small number of discrete steps. This excludes the construction of arbitrary guided tours in virtual space, since it is not immediately obvious how navigation in virtual space may be properly discretized. In this case, as we will discuss later, a guided tour may be constructed using a programmed agent showing the user around.

For navigation in the concept graph, as well as for the activation of the media presentation gadget, the discretization pre-condition holds, and a guided tour may be composed from a finite number of discrete steps, reflecting the choice of the user for a particular node or interaction with the presentation gadget.

For example, in the abramovic dossier, the user has the option to go from the Main node to either Artworks, Video Installations or Interviews, and from there on further to any of the items under the chosen category. Tracking the actual sequences of choices of a user would suffice to create a guided tour, simply by re-playing all steps.

To obtain more interesting tours, we may track the navigation behavior of several experts for a particular task, for example retrieving information about the installation Terra degli della Madre. In case the experts disagree on a particular step in the tour, we may take the majority decision, and possibly correct this by adjusting the weight for one or more experts. When we have a database of tours from a number of experts, we may offer the user a choice of tours, and even allow to give priority to one or more of his/her favorite experts, again simply by adjusting the weighting scheme.

As a technical requirement, it must be possible to normalize interaction sequences, to eliminate the influence of short-cuts, and to allow for comparison between a collection of recordings. For the actual playback, as a guided tour, a decision mechanism is needed that finds the advice at each decision point, from each expert, to select the best step, according to a decision rule that takes the weighting scheme into account.

PERSONALIZATION BY EXPERT RATING

Now assume that we have a weight function w, that assigns to each expert e_{i}$a\; weight\; \$0\; \backslash leq\; \%b\_\{i\}\; \backslash leq\; 1$, indicating the relevance of expert i. Then for a particular node n we may assume to have an advice $\%a\_\{i\}\; =\; x$, with weight $\%b\_\{i\}$ and x in $S(n)$. If an expert has no advice for this node, we may simply assume its weight to be 0. For a collection of experts, the final advice will be $\%a(n)\; =\; \%a\_\{i\}(n)$ with weight $\%b\_\{i\}$ and $w(e\_\{i\})\; >\; w(e\_\{j\})$ for $i\; !=\; j$. If no such advice $\%a\_\{i\}(n)$ exists, we may query the user to decide which expert has preference, and adapt the weights for the experts accordingly. This procedure can be easily generalized to nodes $n\_\{\; \backslash overline\{p\}\; \}$ with history $\backslash overline\{p\}$.

To cope with possible shortcuts, for example when a choice is made for a node at three levels deep, we must normalize the path, by inserting the intermediate node, in order to allow for comparison between experts.

Now assume that we have expert navigation paths with cycles, for example $n\; \_\{\; \backslash overline\{p\}\; ->\; n\; \_\{\; \backslash overline\{p\}1\}\; ->\; n\; \_\{\; \backslash overline\{p\}13\}$, where actually $n\; \_\{\; \backslash overline\{p\}\; =\; n\; \_\{\; \backslash overline\{p\}13\}$, which happens when we return to the original node. In general such cycles should be eliminated, unless they can be regarded as an essential subtour. However, in this case, they could also be offered explicitly as a subtour, if they have length $>=\; 4$. When offering guided tours for which several variants exist, we may allow the user to simply assign weights to each of the experts from which we have a tour, or allow for incrementally adjusting the weight of the experts, as feedback on the actual tour presented.

INTELLIGENT GUIDANCE -- REALIZATION

• filtering the information space according to the user's perspective, and
• intelligent agents, that (pro) actively aid the user in searching the information space.

Intelligent agents is an approach stemming from artificial intelligence which allows for providing guidance in a variety of ways, possibly even in an embodied form using a face or humanoid figure to give suggestions to the user on what interactions to perform, an approach that we will discuss later on.

For selecting the items to be presented in a guided tour, the most obvious way is to pre-define a sequence based on user profiles. Very likely this can be done in a more flexible way in a rule-based manner, applied to a template tour. More interesting, however, is to generate guided tours dynamically based on tracking actual user interaction of (expert) users, using techniques from prediction theory, as explained in the previous sections.

...

---

Fig. 4: Construction tool

---

A special case of a guided tour is the tool environment constructed for the Revolution installation of Jeffrey Shaw, which allows for experimenting with the (de-) construction of the installation, fig. 4, and exhibition parameters, fig. 5.

...

---

Fig. 5: (a) Light	(b) Material

---

Tracking interaction with such 3D models is, given the limitations imposed by the tool environment, relatively simple, and can be used for creating a repository of navigation sequences. More difficult, however, is to find proper normalizations for these interactions, and so in this case we may possibly have to rely on expert weighting only.

agent technology

In  [Hoorn et al. (2004)] we have investigated the use of embodied agents in a digital dossier for the artist Marinus Boezem, fig. 6. To allow for a discrete mode of navigation we have used a map, displaying the interesting parts of the atelier, which contains locations where relevant information can be obtained, such as a filmprojector, for displaying interviews, a cabinet that contains biographical material and textual descriptions of the artworks, and an exhibition environment that displays (3D models of the) artworks. To construct a guided tour, we deployed a humanoid agent that shows the user around.

...

---

Fig 6. Overview	Interview	Agent

---

In a user evaluation test we found that humanoid agents where instrumental in providing information about the re-installation of artworks, but interestingly also that believability was positively affected by the degree of realism of the agent,  [Van Vugt et al. (2006) ]. However, in creating guided tours for the current generation of digital dossiers, using concept graphs for navigation instead of a spatial metaphor, we will not use humanoid agents. Our agent technology, however, can be used in a fruitful way.

In the I-GUARD project (Intelligent Guidance in Archives and Dossiers). we investigate how to realize advice functions, implemented using agent technology,  [Eliens et al. (2002)], based on actual navigation paths obtained by tracking expert users, that offer the user at any navigation point a choice of continuations and/or a selection of guided tours, focussing on a topic of interest.

INCREMENTAL ADAPTATION OF RECOMMENDATIONS

In the CHIP project (Cultural Heritage Information Personalization), the aim is to develop a recommender system that generates a collection of artworks in accordance with the users' preferences based on the rating of a small sample of artworks. The properties on which the recommendation is based include period, artist, and genre. The recommender system will also be used to generate guided tours, where apart from the already mentioned properties the location (the proximity in the actual museum) will be taken into account.

Using a weighting scheme on the properties, that is a difference metric on the properties, a graph can be created, giving a prioritized accessibility relation between each artwork and a collection of related artworks. By changing the weight for one of the properties, for example location, in case the tour is generated for the actual museum, the priority ordering may be changed, resulting in a different tour.

In contrast to the successor function for nodes in the concept graph of the digital dossier, we may assume to have a weighted successor function $S\_\{w\}(n)\; =\; \{\; (n\_\{1\},\%w\_\{1\}),...,(n\_\{k\},\%w\_\{k\})\; \}$, with $\%w\_\{i\}\; =\; w(n\_\{i\})$ the weight defined by the relevance of the node $n\_\{i\}$, with respect to the attributes involved. In a similar way as for the digital dossier, user tracking may be deployed to incrementally change the weight of the arcs of the graph, reflecting the actual preference of the user when deviating from an existing guided tour. In the remainder of this paper we will give the outline of a recommender model supporting the incremental adaptation of preferences by user feedback.

SERIAL RECOMMENDER MODEL

Admittedly not the best way to do research, although common practice, we found a good starting point for modelling recommender systems, by googling on serial recommender, in a paper from Microsoft Research on privacy in distributed recommender systems,  [Oard et al. (2006)]. The model introduced in  [Oard et al. (2006)], distinguishes between:

recommender model

  $U\; =\; user$
  $I\; =\; item$
  $B\; =\; behavior$
  $R\; =\; recommendation$
  $F\; =\; feature$
  

and allows for characterizing observations (from which implicit ratings can be derived) and recommendations, as follows:

• observations -- $U\; \backslash *\; I\; \backslash *\; B$
• recommendations -- $U\; \backslash *\; I$

In a centralized approach the mapping $U\; \backslash *\; I\; \backslash *\; B\; ->\; U\; \backslash *\; I$ provides recommendations from observations, either directly by applying the $U\; \backslash *\; I\; ->\; I\; \backslash *\; I$ mapping, or indirectly by the mapping $U\; \backslash *\; I\; ->\; U\; \backslash *\; U\; ->\; I\; \backslash *\; I$, which uses an intermediate matrix (or product space) $U\; \backslash *\; U$ indicating the (preference) relation between users or user-groups. Taken as a matrix, we may fill the entries with distance or weight values. Otherwise, when we use product spaces, we need to provide an additional mapping to the range of $[0,1]$, where distance can be taken as the dual of weight, that is $d\; =\; 1\; -\; w$.

In a decentralized approach,  [Oard et al. (2006)] argue that it is better to use the actual features of the items, and proceed from a mapping $I\; \backslash *\; F\; ->\; U\; \backslash *\; I\; \backslash *\; R$. Updating preferences is then a matter of applying a $I\; \backslash *\; B\; ->\; I\; \backslash *\; F$ mapping, by analyzing which features are considered important.

For example, observing that a user spends a particular amount of time and gives a rating r, we may apply this rating to all features of the item, which will indirectly influence the rating of items with similar features.

  B = [ time = 20sec, rating = r ]
  F = [ artist = rembrandt, topic = portrait ]
  R = [ artist(rembrandt) = r, topic(portrait) = r ]
  

 [Oard et al. (2006)] observe that B and R need not to be standardized, however F must be a common or shared feature space to allow for the generalization of the rating of particular items to similar items.

With reference to the CHIP project, mentioned in the previous section, we may model a collection of artworks by (partially) enumerating their properties, as indicated below:

  A = [ $p\_\{1\},\; p\_2$, ... ]
  where $p\_\{k\}$ = [ $f\_1\; =\; v\_1$, $f\_2\; =\; v\_2$, $...$ ]
  

with as an example

  $A\_\{nightwatch\}$ = [ artist=rembrandt, topic=group ]
  $A\_\{guernica\}$ = [ artist=picasso, topic=group ]
  

Then we can see how preferences may be shared among users, by taking into account the (preference) value adhered to artworks or individual properties, as illustrated in fig. 7.

...

---

Fig 7. Users, artworks and properties

---

As a note, to avoid misunderstanding, Picasso's Guernica is not part of the collection of the Rijksmuseum, and does as such not figure in the CHIP studies. The example is taken, however, to clarify some properties of metrics on art collections, to be discussed in the next section.

CONTENT METRICS

To measure similarity, in information retrieval commonly a distance measure is used. In mathematical terms a distance function $d:X->[0,1]$ is distance measure if:

distance metric

       $d(x,y)\; =\; d(y,x)$
       
       $d(x,x)\; =\; 0$
  

From an abstract perspective, measuring the distance between artworks, grouped according to some preference criterium, may give insight in along which dimesnion the grouping is done, or in other words what attributes have preference over others. When we consider the artworks

    $a\_1$ = [ artist = rembrandt, topic = self-portrait ]
    $a\_2$ = [ artist = rembrandt, name = nightwatch ]
    $a\_3$ = [ artist = picasso, topic = self-portrait ]
    $a\_4$ = [ artist = picasso, name = guernica ]
  

we may, in an abstract fashion, deduce that if then , however if the reverse is true, that is then . Somehow, it seems unlikely that $a\_2$ and $a\_4$ will be grouped together, since even though their topic may considered to be related, the aesthetic impact of these works is quite different, where $self\; portrets$ as a genre practiced over the centuries indeed seem to form a 'logical' category. Note that we may also express this as if we choose to apply weights to existing ratings, and then use the observation that if then to generate a guided tour in which $a\_3$ precedes $a\_2$.

For serial recommenders, that provide the user with a sequence of items $...,\; s\_\{n-1\},\; s\_\{n\},\; ...$, and for $s\_\{n\}$ possibly alternatives $a\_1,\; a\_2,\; ...$, we may adapt the (implied) preference of the user, when the user chooses to select alternative $a\_\{k\}$ instead of accepting $s\_\{n\}$ as provided by the recommender, to adjust the weight of the items involved, or features thereof, by taking into account an additional constraint on the distance measure. Differently put, when we denote by $s\_\{n-1\}\; |->\; s\_\{n\}/[\; a\_1,\; a\_2,\; ...\; ]$ the presentation of item $s\_\{n\}$ with as possible alternatives $a\_1,\; a\_2,\; ...$, we know that for some k, if the user chooses for $a\_\{k\}$ In other words, from observation $B\_\{n\}$ we can deduce $R\_\{n\}$:

  $B\_\{n\}$ = [ time = 20sec, forward = $a\_\{k\}$ ]
  $F\_\{n\}$ = [ artist = rembrandt, topic = portrait ]
  $R\_\{n\}$ = [  ]
  

leaving, at this moment, the feature vector $F\_\{n\}$ unaffected. Together, the collection of recommendations, or more properly revisions $R\_\{i\}$ over a sequence S, can be solved as a system of linear equations to adapt or revise the (original) ratings. Hence, we might be tempted to speak of the R4 framework, rate, recommend, regret, revise. However, we prefer to take into account the cyclic/incremental nature of recommending, which allows us to identify revision with rating.

MEASURES FOR FEEDBACK DISCREPANCY

So far, we have not indicated how to process user feedback, given during the presentation of a guided tour, which in the simple case merely consists of selecting a possible alternative. Before looking in more detail at how to process user feedback, let us consider the dimensions involved in the rating of items, determining the eventual recommendation of these or similar items. In outline, the dimensions involved in rating are:

• positive vs negative
• individual vs community/collaborative
• feature-based vs item-based

Surprisingly, in  [Wang et al. (2007)] we found that negative ratings of artworks had no predictive value for an explicit rating of (preferences for) the categories and properties of artworks. Leaving the dimension individual vs community/collaborative aside, since this falls outside of the scope of this paper, we face the question of how to revise feature ratings on the basis of preferences stated for items, which occurs (implicitly) when the user selects an alternative for an item presented in a guided tour, from a finite collection of alternatives.

A very straightforward way is to ask explicitly what properties influence the decision. More precisely, we may ask the user why a particular alternative is selected, and let the user indicate what s/he likes about the selected alternative and dislikes about the item presented by the recommender. It is our expectation, which must however yet be verified, that negative preferences do have an impact on the explicit characterization of the (positive and negative) preferences for general artwork categories and properties, since presenting a guided tour, as an organized collection of items, is in some sense more directly related to user goals (or educational targets) than the presentation of an unorganized collection of individual items. Cf.  [Van Setten (2005)].

So let's look at $s\_\{n-1\}\; |->\; s\_\{n\}/[\; a\_1,\; a\_2,\; ...\; ]$ expressing alternative selection options $a\_1,\; a\_2,\; ...$ at $s\_\{n\}$ in sequence $S\; =\; ...,\; s\_\{n-1\},\; s\_\{n\}$. We may distinguish between the following interpretations, or revisions:

• neutral interpretation -- use
• positive interpretation -- increase $w(feature(a\_\{k\}))$
• negative interpretation -- decrease $w(feature(s\_\{n+1\}))$

How to actually deal with the revision of weights for individual features is, again, beyond the scope of this paper. We refer however to  [Eliens (2000)], where we used feature vectors to find (dis)similarity between musical fragments, and to  [Schmidt et al. (1999)], on which our previous work was based, where a feature grammar is introduced that characterizes an object or item as a hierarchical structure, that may be used to access and manipulate the component-attributes of an item.

CONCLUSIONS

In this paper we have shown how to adapt guided tours based on tracking expert users by modifying the weights attached to the experts that contributed to the construction of this tour. The application of these techniques requires that choices are discrete and hence do not apply to arbitrary navigation in virtual environments, unless we find proper ways to encode such navigation as a small finite collection of discrete steps. Also in the discrete case, however, we must be able to normalize navigation paths, in order to compare and weigh the contribution of the experts involved.

We have generalized our approach to a wider class of serial recommenders, and indicated how to apply the revision of ratings in an incremental fashion to adapt an existing tour to personal preferences, reflecting the actual navigation behavior of users.

As future work, we wish to investigate how we can use both positive and negative user feedback to revise and refine ratings for the actual features involved. Additionally, we would like to study features not directly related to artworks, but for example to group norms or personal likes and dislikes.

ACKNOWLEDGEMENT(S)

We thank the reviewer who valued our original submission, which was meant to be a short paper, so positively that we were invited to write an extended paper. We took the challenge and hopefully not to the disappointment of our anonymous reviewer.

AUTHOR BIOGRAPHY

ANTON ELIENS

is coordinator of the multimedia curriculum at the computer science department of the Faculty of Sciences of the Vrije Universiteit Amsterdam, He has written numerous papers and published books on distributed logic programming and object oriented software engineering.

YIWEN WANG

is Ph.D. researcher in the CHIP project from the Technische Universiteit Eindhoven. She does her research at the Rijksmuseum Amsterdam, together with the other members of the CHIP team.

REFERENCES

[CHIP] Aroyo, L., Rutledge, L., Brussee, R., de Bra, P.,

Gorgels, P., Stash N., Veenstra, M. (2005), Personalized Presentation and Navigation of Cultural Heritage Content, Multimedia and Expo, ICME 2005. IEEE International Conference, 2005

[Prediction] Cesa-Bianchi N. and Lugosi G. (2006),

Prediction, Learning, and Games, Cambridge University Press

[OO] Eliens A. (2000),

Principles of Object-Oriented Software Development, Addison-Wesley Longman, 2nd edn.

[Platform] Eliens A., Huang Z., and Visser C. (2002),

A platform for Embodied Conversational Agents based on Distributed Logic Programming, In: Proc. AAMAS Workshop -- Embodied conversational agents - let\'s specify and evaluate them!, Bologna 17/7/2002

[Navigate] Eliens A., van Riel C., Wang Y. (2006),

Navigating media-rich information spaces using concept graphs -- the abramovic dossier, In Proc. InSciT2006, 25-28 Oct. 2006, Merida, Spain

[Agents] Hoorn J., Eliens A., Huang Z., van Vugt H.C.,

Konijn E.A., Visser C.T. (2004). Agents with character: Evaluation of empathic agents in digital dossiers, In: Proc. AAMAS 2004, Emphatic Agents, New York 19 July - 23 July, 2004

[Privacy] Oard D.W., Leuski A. and Stubblebine S. (2006),

Protecting the Privacy of Observable Behavior in Distributed Recommender Systems, In: Proc. SIGIR 2003

[Guide] Riel C. van, Eliens A., Wang Y. (2006),

Exploration and guidance in media-rich information spaces: the implementation and realization of guided tours in digital dossiers, In: Proc. InSciT2006, 25-28 Oct. 2006, Merida, Spain

[Maps] Riel C. van, Wang Y. and Eliens A. (2006b),

Concept map as visual interface in 3D Digital Dossiers: implementation and realization of the Music Dossier, In: Proc. CMC2006, Costa Rica, Sept 5-8 2006

[Cones] Robertson G.G. and MacKinlay J.D. (1991),

Cone trees: animated 3D visualizations of hierarchical information, In: Proc. of the SIGCHI 1991, pp. 189-194

[Features] Schmidt, A.R. Windhouwer M.A., Kersten M.L,

(1999). Indexing real-world data using semi-structured documents, CWI Report INS-R9902

[MMDBMS] Subrahmanian V.S. (1998),

Principles of Multimedia Databases, Morgan Kaufmann

[Hybrid] Van Setten M. (2005),

Supporting People in Finding Information -- Hybrid recommender Systems and Goal-based Structuring, Ph.D. Thesis, Telematica Institute Netherlands

[Experiment] Van Vugt, H. C., Konijn, E. A., Hoorn, J. F., Keur, I.,

& Eliens, A. (2006). Realism is not all! User Engagement with Task-Related Interface Characters, Interacting with Computers, 2006

[Present] Wang Y., Eliens A., van Riel C. (2006),

Content-oriented presentation and personalized interface of cultural heritage in digital dossiers, In: Proc. InSciT2006, 25-28 Oct. 2006, Merida, Spain

[User] Wang Y., Aroyo L., Stash N., Rutledge L. (2007),

Interactive User Modeling for Personalized Access to Museum Collections -- The Rijksmuseum Case Study, accepted for 11th Conf. on User Modeling, UM 2007, June 25-29, Corfu, Greece

---

[] readme course preface I 1 2 II 3 4 III 5 6 7 IV 8 9 10 V 11 12 afterthoughts appendix references examples resources _

---

(C) A. Eliëns 21/5/2007

You may not copy or print any of this material without explicit permission of the author or the publisher.
In case of other copyright issues, contact the author.