Handbook of Multimedia for Digital Entertainment and Arts- P22 pps

YouTell: A Web 2.0 Service for Community Based Storytelling How to apply storytelling for professional communities can be enabled by Web 2.0and Social Software.. The suc-cessful realizat

Fig 3 A comparison of the existing storytelling platforms

to drive on it The system can be interpreted as community of practice (drivers whohave access to the Internet via mobile phones or PDA’s) and collaborative, since it

is very important to get real time feedback from the users

Figure3presents a summary of all systems presented Features presented in thetable are very important for one storytelling system nowadays to meet all the re-quirements of the users

YouTell: A Web 2.0 Service for Community Based Storytelling

How to apply storytelling for professional communities can be enabled by Web 2.0and Social Software We have designed and developed youTell using Web 2.0 ser-vice for community based storytelling It is based on a social software architecture

called Virtual Campfire.

Virtual Campfire

In order to make knowledge sharing a success for any kind of professional munity, independent of size or domain of interest, a generic community enginefor Social Software is needed After some years of experience, with the sup-port of professional communities two different products emerged: a new reflectiveresearch methodology called ATLAS (Architecture for Transcription, Localization,

com-and Addressing Systems) [10] and a community engine called LAS (LightweightApplication Server) [23] The research challenge in ATLAS was to incorporate thecommunity members as stakeholders in the requirements and software engineeringprocess as much as possible In the end, all community design and engineering ac-tivities should be carried out by the community members themselves, regardless oftheir technical knowledge While this ultimate goal of taking software engineers out

of the loop is rather illusionary in the moment, we have targeted realizing a genericarchitecture based on the research methodology It allows community members tounderstand their mediated actions in community information systems In its reflec-tive conception the community information systems based on ATLAS are tightlyinterwoven with a set of media-centric self monitoring tools for the communities.Hence, communities can constantly measure, analyze and simulate their ongoingactivities Consequently, communities can better access and understand their com-munity need This leads to a tighter collaboration between multimedia communityinformation systems designers and communities Within UMIC we have developedthis complex scenario of a mobile community based on our real Bamiyan Develop-ment community, and the ATLAS/LAS approach Virtual Campfire is an advancedscenario to create, search, and share multimedia artifacts with context awarenessacross communities Hosted on the basic component the Community Engine, VirtualCampfire provides communities with a set of Context-Aware Services and Multi-media Processor Components to connect to heterogeneous data sources Throughstandard protocols a large variety of (mobile) interfaces facilitate a rapid design andprototyping of context-aware multimedia community information systems The suc-cessful realization of a couple of (mobile) applications listed as follows has provedthe concept and demonstrated Virtual Campfire in practices: MIST as a multimediabased non-linear digital storytelling system; NMV as a MPEG-7 standard basedmultimedia tagging system; (Mobile) ACIS as a Geographic Information System(GIS) enabled multimedia information system hosting diverse user communities forthe cultural heritage management in Afghanistan; and finally CAAS as a mobileapplication for context-aware search and retrieval of multimedia and communitymembers based on a comprehensive context ontology modeling spatial, temporal,device and community contexts All these applications employ the community en-gine and MPEG-7 Services within the Virtual Campfire framework Other servicesand (mobile) interfaces are applied according to different communities require-ments Virtual Campfire is running on Wireless Mesh Networks to apply high andstable network data transfer capability, and low cost, in developing countries

In order to use Web 2.0 feature, related community concepts for storytelling, aprototype called YouTell has been developed within the Virtual Campfire scenario.Figure 4 gives an overview of this new web service Additionally to storytellingfunctionality an expert-finding service is integrated Web 2.0 techniques as taggingand giving feedback contribute to a comprehensive role model for storytelling too.Tags can be analyzed for a dynamic classification of experts This role model is alsoused to represent the behavior and influence of every user

In our previous research, we have focused on how to generate stories by applying

the Movement Oriented Design (MOD) paradigm, which divides stories into Begin,

Fig 4 An overview on the YouTell concepts

Middle, and End parts [21] We have designed and deployed a so-called MultimediaIntegrated Story-Telling system (MIST) to create, display, and export non-linearmultimedia stories [26] MIST proves to be applied in domains of cultural heritagemanagement well, in order to organize a great amount of multimedia content re-lated to a monument or a historical site MIST can also be used as an e-learningapplication to manage multimedia learning stuff or as an e-tourism support system

to generate personalized tour guide

Drawbacks or missing features are also exposed during the deployment First ofall, MIST lacks the mechanism to support users’ collaborative storytelling explic-itly That means, more than one users are able to work on the same story together,while their activities are not recorded for each user respectively but mixed Second,MIST can be used to create and view stories But users can not give any personalcomments to the stories Third, it is almost impossible to search stories in the largestory repositories, since these multimedia stories do not possess proper metadata todescribe it content Finally, MIST lacks authority, if a story has a serious usage e.g.learning knowledge in a certain area The question arises, who are the experts in thestorytelling communities and have more potentials to create arts?

YouTell enables communities to have joint enterprises (i.e story creation), tobuild a shared repertoire (i.e stories) and to engage mutually (i.e expert contacts).Therefore, YouTell build a platform for a community-of-practice with a number

of experts [29] YouTell has also employed the most highlighted Web 2.0 features

like tagging and feedback from amateur Hence, the conflict between experts and

amateur is dealt wish in YouTell.

The main design concepts as well as algorithms of the YouTell system are anappropriate role model as well as user model for storytelling, Web 2.0 tagging fea-tures, the profile-based story search algorithm, and expert finding mechanism.

The Role Model

All roles which should be taken into account for storytelling is specified in YouTell(cf Figure5) A new YouTell user John Doe gets necessary rights to execute basicfeatures like tagging, viewing, rating and searching for a story

Experts are users which have the knowledge to help the others There exist three

different sub roles A YouTell technician can aid users with administrative questions.

A Storyteller knows how to tell a thrilling story And finally a Maven is characterized

as possessing good expertise A user has to give a minimum number of good advices

to the communities in order to be upgraded to an expert

Administrators have extended rights which are necessary for maintenance issues.

The system admin is allowed to change system and configuration properties Story

sheriffs can delete stories and media Additionally, there exists the user admin He

manages YouTell users and is allowed to lock or delete them

A producer create, edit and manage stories The producer role is divided into the production leader who is responsible for the story project, the author who is responsible for the story content, the media producer who is responsible for used media, the director who is responsible for the story, and finally the handyman who

is a helper for the story project

Fig 5 The YouTell role model

The role called Bandits classifies users which want to damage the system cording to their different behavior, they can be a troll, a smurf, a hustler or a

Ac-munchkin.

In contrast to bandits there exist two prestige roles: the connector and the domain

lord Whereas the connector knows many people and has a big contact network,

the domain lord both has a great expertise and, at the same time, is an excellent

storyteller

Web 2.0 for Storytelling: Tagging and Rating

If a YouTell user wants to create a story, he first has to create a story project Withregard to his wishes he can invite other YouTell members to join his project Ev-ery team member is assigned to at least one producer role Every YouTell user cantag stories to describe the related content Because the widely-in-use plain taggingapproach has several disadvantages [14], a semantic tagging approach is used, too.Besides, users’ rating and viewing activities on stories are also recorded As de-picted in Figure 6, A YouTell story are described with tags, rated by users Thepopularity is also reflected by the viewing times

Profile-based Story Searching

In comparison to MIST, YouTell has enhanced the story searching feature greatly.Additionally to a content based search by title or tags, a profile based search isoffered to users Figure7shows how the profile based story search works

In the following the corresponding algorithm is explained in detail The set ofall stories, which haven’t been seen and created by the user is described trough

S D fS1; :::; Sng The function  W S 7! W L assigns a set of tags to a story, R isthe set of all ratings, RSi is set set of ratings of story Si; Si2 S

Fig 6 Information board of a YouTell story

Fig 7 Profile based story search algorithm

Input of the algorithm is a user made tag list W D fsw1; :::swkg Additionallyfurther information are necessary: the maximal result length n and the set of storyratings B of user with a similar profile For computation of these users the Pearson

r algorithm is taken(cf [28])

Considered are user with similar or opposite ratings If the ratings are similar thePearson value is near to 1, if they are opposite the value is sear to -1 In the firstcase stories with similar ratings, in the second case stories with opposite ratings arerecommended The value has to be in a threshold L to be suitable The Pearson value

is computed with the following formula:

wa;b D

Pm iD1.ra;i ra/  rb;i rb/

i D1.ra;i  ra/2Pm

i D1.rb;i rb/2

The Pearson value between user profile a and compared profile b is represented

through wa;b The variable m corresponds to the story count, i is a particular storyand r its rating The average ratings of profile a is displayed through ra

It holds B D fBS 1;    ; BSkg with Si 2 S; 1  i  k Furthermore BSicorresponds to the set of story ratings of user with a similar profile for story Siandfinally it holds .RS i/ D BS i

1 step:

Group the stories: The first group G1 corresponds to the story set S1;    ; Sm,

Si 2 S with W  .Si/ and .RSi/ 2 B The second group G2 contains thestories S1;    ; Sl, Sj 2 S with .Sj/ \ W ¤ ;; W ª .Sj/ and .RSi/ 2 B

2 step:

1 Take group G1D fS1;    ; Smg

a Compute the story ratings median BSi for every story Si

b Build a ranking corresponding to the medians

2 If jG1j < n, take group G2 D fS1;    ; Slg

a Be P W S 7! R a function, which assigns a number of points to every story

b For every j; 1  j  l it holds P Sj/ D 0

c For every tag swi

For every storySj

If swi 2 .Sj/:

Compute the median mj of ratings BSjMap the result to the range [1,5]: m0j D mj C 3

P Sj/C D m0j

d Sort the stories by their score

3 Build an overall ranking with the rankings from group 1 and group 2 This ing is the output of the algorithm

rank-Expert Finding System

Users who have questions can contact an expert A special algorithm and useful userdata are necessary to determine the users knowledge, in order for the users to findthe best fitting expert

For every user there exists a user profile which contains the following information:

Story data are generated when a user visits or edits a story.

Expert data are created with given/ received expert advices

Personal data represent the user knowledge the user has acquired in the real world.

These data are typed in by the user itself

With these information three tag vectors are created They will be weighted summed

up and normalized Such a vector has the following form:

264

t aga val uea

t agb valueb

t agc valuec

     

375

The final value of each tag represents the users knowledge assigned to the relatedtag A value near to zero implies that the user only knows few, where as a value near

1 implies expertise at this topic

Now it will be described how the data vector is composed First the story datavector will be created For every story a user has visited and for every story for which

the user is one of the producer, the corresponding story/media tags will be stored in a

vector The respective value is computed with the formula value D AV  DV  BF and

– AV OD count of appearances of a tag

– DF OD date factor: The older a date, the more knowledge is lost The valuelies between 0 and 1 A 1 stands for an actual date, a zero for a very oldone Four weeks correspond to a knowledge deficit of 5 percent It holds:

DF D 1  b#weeks4 c  0:05/

– BF OD rating factor: This value is computed by the explicit and implicit feedbackwhich has been given

Then the story data vector d is computed:

d D Story visit vector  0:35 C Story edit vector  0:65:

After that a normalization to the range Œ0; 1 will be done: Let S D fs1; :::; sng be

the set of all tags, which occur within the set of data vectors and let v.s/ be the

corresponding value

8s 2 S v.s/normD v .s/  v min

v max  v min and v min D minfv.s1/; :::; v.sn/g; v max D maxfv.s1/; :::; v.sn/g:

In a second step the expert data vector is computed For every expert advice auser has given/ obtained the corresponding tags are stored in a vector The respectivevalue will be calculated analogously to the above computation and it holds:

expert data vector D advicegiven 0:8 C ad vi ceobtained 0:2:

Third the personal data vector is computed With the information the system gotfrom the user tags and its corresponding values will be obtained These will be takenfor this vector

In a last step the final vector will be computed:

data vector D 0:4  expert data vector C 0:4  story data vector

C0:2  personal data vector:

To find an expert first a vector v D fs1; w1; :::; sm; wmg will be created with thetags the user has specified Then this vector will be compared with all existing data

vectors w1;    ; wn The user with the best fitting vector will be the recommendedexpert

The vectors have the following form:

z D s1; w1; s2; w2;    ; sm; wm/, whereas si is the i -th tag and wi the sponding value

corre-1 Repeat for every vector wj; 1  j  n

3 Repeat for every tag si; 1  i  m of vector v

4 If si D sj k, sj k 2 wj: diffj D diffj C wi wj k/

5 else diffj D diffj C 1

Output of this algorithm is the user for which data vector u holds: u D wj mitdiffj D minfdiff1;    ; diffng

Web 2.0 for the Expert-finding Algorithm

How does Web 2.0 features like tagging and esp feedback influence on finding? Users can give feedback to stories and for expert advices Feedback is veryimportant for YouTell, because it delivers fundamental knowledge for executing theprofile based search and defining the user’s expert status Furthermore the visu-alization of feedback results (i.e average ratings, tag clouds) help user to get animpression of the experts/story’s quality

expert-Explicit and implicit feedback techniques are used After visiting a story tively getting an expert advice the user has the possibility to fill out a questionnaire.This explicit form of giving feedback is fundamental for YouTell But not every userlikes filling out questionnaires [31] Therefore, also implicit feedback is employed.Although this is not as accurate as explicit feedback, it can be an effective substitute[31] In YouTell the following user behavior will be considered: The more one uservisits one story the more interesting it is The more a story is visited by all users, themore popular it is

respec-In addition, the integrated mailbox service offers the possibility to handle allnecessary user interaction of the YouTell community Users need to send mes-sages when they want to ask an expert, give an expert advice, invite a new teammember, etc

Implementation of the YouTell Prototype

An overall architecture of YouTell is illustrated in Figure8 YouTell is realized asclient/server system and is integrated in the LAS system [25] implemented in Java.The client, implemented as a web service, communicates via the HTTP protocolwith the las server by invoking service methods The LAS server handles the usermanagement and all database interactions New services like the expert, mailbox,YouTell user and storytelling service extend the basic LAS features and fulfill allfunctionality needed by YouTell

The story service extends already existing MIST features and includes methodsfor the management of story projects and searching for stories The expert ser-vice contains functions for computation and management of the expert data vectors.The mailbox service manages the mailbox system The YouTell user service extendsthe LAS user service and offers the possibility to add and edit user specific data

Fig 8 System architecture of YouTell

In addition, YouTell needs several different servers to work properly The clientsystem communicates via the HTTP protocol with an Apache tomcat server TheirServlets and JSPs are executed for the user interface of YouTell (cf Figure9) InYouTell the storytelling board is integrated with Java applets which run on the client.All media of the YouTell community are stored on a FTP server The communicationwith the used databases (eXist and DB2) is realized by the LAS server

Fig 9 YouTell screenshots and functionality description

YouTell Evaluation

The evaluation of YouTell consists of three parts First users had the possibility totest the YouTell prototype The results are described in Sub Section After that thealgorithms for profile based search (Sub Section29) and for expert finding (section29) are evaluated The algorithms need a great amount of data to work properly.Because the available data were not sufficient, test data had to be generated for aconvincing system evaluation result

Figure10shows how often LAS services have been called From the statistics, itshows that the story service has been used most frequently and expert service callstook the longest time

Fig 10 Calling statistics

Fig 11 Questionnaire results

The questionnaire has been completed by 10 persons within a test session.Figure 11shows some results The worst rating was given for the loading times.Additionally, YouTell is not user friendly enough YouTell as a whole has been ratedwith an average score of 2.1 and corresponds to the German school grade ”good”

To sum up the results, YouTell has been accepted by the test user but has to beimproved

Profile Based Story Search

To evaluate the profile based story search an approach analogously to the proceeding

of Shardanand and Maes ([28]) has been performed

1 Delete 20 percent of story visits of an arbitrary chosen user U

2 Run profile based story search without specifying tags Compare results and moved stories Store percentage of coverage

re-Table 3 Evaluation result: achieved hit rate (in pro cent)

Minimum Maximum Mean Standard variance Profile based search with keywords 60 100 97.11 8.237

Profile based search without keywords 60 100 92.69 12.445

3 Run profile based story search with specifying tags that where assigned to thedeleted story set Compare results and removed stories Store percentage ofcoverage

The evaluation results are shown in Table3 The average hit rate of 97,11 cates that not all removed stories are found There exists a simple reason for this.Only stories visited by similar users can occur in the result set Because all removedstories haven’t been visited by similar users, the algorithm delivers the exactly rightresults

indi-Also in step 3 of the evaluation proceeding not all stories are found But thiswas expected and the hit rate corresponds to the analysis results of Shardarnand andMaes [28]

Expert Finding Algorithm

The expert finding algorithm delivers user/tag pairs with an expert value lyingbetween one and zero To evaluate the algorithm the value distribution has beenanalyzed

In Figure12the distribution of the expert values is depicted For every number

on the x- axis the frequency of user/tag pairs with an appropriate expert value isdenoted Figure13shows the same values separated by the singular tags

So both figures show that the expert knowledge distribution is approximately mally distributed Because the test data were predominantly normally distributedthis result was expected In Figure12the expert value 1 has a peak which seems

nor-to be unusual at first glance This can be explained by the used normalization: ter computing the data vectors they will be normalized in the range from 0 to 1separated by the tags Therefore, for every tag exists a user/tag pair with the value

Af-1 resp 0 This approach establishes the possibility to represent the knowledge signed to particular a tag within the YouTell community Figure13shows that thedistinct knowledge function differ This implies that the knowledge about particulartopics is differently pronounced within the YouTell test community

as-In addition to classification of the users’ knowledge, the expert finding algorithmdelivers a measurement for analyzing the community knowledge

Fig 12 Distribution of expert knowledge

Fig 13 Distribution of expert knowledge according to keywords

Summary

In this chapter, storytelling is discussed as a new means of creating arts based

on Web 2.0 features and communities of practice We illustrate a use scenario todemonstrate why and how storytelling is useful The related work about the taggingapproach and storytelling platforms etc is discussed in Section29 Section29per-tains to the design and main features of the community-based storytelling system

YouTell Section29introduces the implementation of prototype YouTell tion results based on hands-on experiences from the user communities are presented

Evalua-in Section

In summary, we combine Web 2.0 and communities of practice with expert ing in a storytelling platform YouTell to create arts YouTell is featured with a rolemodel for storytelling system, the tagging concepts, the profile based story searchapproach, and the expert finding mechanism The YouTell architecture is discussedtogether with the prototype implementation issues The prototype evaluation resultsshow that the usefulness and performance in profile based story search as well asexpert finding mechanism Generated stories have been further applied to createeducational games in order to train the professional communities [24] Besides con-ceptual approaches and technical realization, the First International Workshop onStory-Telling and Educational Games (STEG’08) has been organized as an annualevent to bring researcher communities on storytelling together

find-In the ongoing future research, new questions arise How can diverse user munities work together seamlessly to create art through Web 2.0 based storytellingapproach? How can amateur be upgraded into experts? The process of the ideasharing can also be carried out in YouTell More application case studies can beexplored How can various Web 2.0 media organized via the YouTell storytellingplatform? We can imagine that a number of Weblog entries or even bookmarks can

com-be organized in a user generated sequence for the storytelling purpose Stories will

be exploited for entertainment with some speech bubbles, so that the expressiveness

of the story narration could be enhanced or some art comments can also be given onthe bubbles (see Fig.14)

Fig 14 Narrating a story with speech bubbles