visual workflow interfaces for editorial processes

Visual workflow interfaces foreditorial processes

Luciano Frizzera

University of Alberta Canada

Milena Radzikowska

Mount Royal University Canada

Geoff Roeder Ernesto Pena and Teresa Dobson

University of British Columbia Canada

Stan Ruecker

IIT Institute of Design USA

Geoffrey Rockwell

University of Alberta Canada

Susan Brown

Guelph University Canada

The INKE Research Group

University of Victoria Canada

AbstractIn this article we provide a discussion of the concept of visual interactive work-flows how they relate to our previous work on structured surfaces and how theyhave been adapted to experiments in managing articles for journal publicationand managing biographical histories being written and tagged in XML We con-clude with a user experience study of the prototypes which suggests that they arerelatively acceptable at the level of reflective response but might benefit frommore iteration in their use of process and element metaphors

1 Introduction

Collaborative scholarly writing and publishingactivities typically involve many different types ofdocuments divided across various stages This is es-pecially true in the case of big projects Documentsmay pile up scattered in many foldersmdashboth realand virtualmdashawaiting tagging revision editingproofreading approval and so on There are work-flow management tools available to help but they

have often been designed with other audiences in

mind so they can be too specialized and somewhat

complex to use The goals of this project were there-

fore to make them less specialized and easier to useThere are several potential advantages of im-

proved workflow management for document pro-

duction and publishing First is the reduction in

general confusion whether by one person or several

about the status of various documents within the

system Since both writing and publishing tend to

Correspondence

Luciano dos Reis Frizzera

Humanities Computing

Office of Interdisciplinary

Studies

1-17 Humanities Centre

University of Alberta

Edmonton Alberta

Canada T6G 2E5

Email

dosreisfualbertaca

Literary and Linguistic Computing Vol 28 No 4 2013 The Author 2013 Published by Oxford University Press onbehalf of ALLC All rights reserved For Permissions please email journalspermissionsoupcom

615

doi101093llcfqt053 Advance Access published on 18 September 2013

at University of V

ictoria on March 12 2015

httpllcoxfordjournalsorgD

ownloaded from

be intermittent tasks with irregular intervals be-

tween work sessions a considerable amount of

time can be spent just in getting back up to speed

on where the work left off Second is the increase inconfidence not necessarily about the status of a

document but about its location among the various

people involved This is particularly important in

cases where the document may be in the hands of

several people simultaneously and it is important to

track how long it is taking in each location so that

bottlenecks can be prevented and the amalgamation

of work carried out efficiently Finally related toreduced confusion increased confidence improved

productivity and other factors there are possible

benefits to the mental and emotional well-being of

the people involvedIn terms of the visual component it is generally

understood that visual representations of data andprocesses can have the advantages of providingan overview of the material insights into theunderlying structure and affordances and benefitsfor non-expert or intermittent users (eg Rueckeret al 2011) However it is also important to avoidthe many potential disadvantages of visualizationBresciani and Eppler (2009) provide a preliminarytaxonomy of gt50 potential cognitive emotionaland social disadvantages which in each case mightbe designer-induced or user-induced

There are several ways in which such improve-ments might be subject to measurement includingcomparisons between using a visual workflow andnot using one for the following

time to create documents publisher hours spent per document published cognitive and affective reports from writers and

editors

This project falls within the larger program ofresearch set out by the Interface Design (ID) teamof Implementing New Knowledge Environments(INKE) and in particular INKErsquos recent work onsupporting the creation and editing of digital ma-terials as opposed to their access and use We havebeen exploring the concept of structured surfaces(Radzikowska et al 2011) to investigate bothvisual and interactive methods to make workflowsmore attractive flexible and useful for scholars We

provide a summary of interactive visual workflowsand how they have been applied and explain whatstructured surfaces are and how the concept can beapplied to workflows

11 WorkflowsMany projects in the digital humanities involveeither digitization or enrichment of existing digitalmaterials The process is usually similar to editorialactivity and can be understood as an ordered check-list or sequence of events or perhaps more accur-ately an ordered graph of events as some can occurin parallel or only after a decision point A re-searcher first enters the raw text then encodesproofs and tests the results There is a process orworkflow to be followed

Visual tools for workflow management thereforeoften also take the form of what is essentially a struc-tured checklist composed of a collection of stageswith transitions including occasional decisionpoints between them In effect they are flowchartsThese tools are widely used in science and business todescribe production process models research in thisarea began as early as the 1970s (eg Aalst 2004)Contemporary examples range from managingearth sciences data at NASA (Berrick et al 2008) toenvironmental performance monitoring (Versteeget al 2006) to modelling business problems(Virine and McVean 2004) There have howeverbeen criticisms both about the lack of visual stand-ards and the poor visual decision choices that resultlsquoOften the construction of a layout for a possiblycomplex workflow is left to the user or the result isvisually unsatisfyingrsquo (Albrecht 2010) The solutionAlbrecht proposes involves better algorithms forautomatic layout (Fig 1)

In the humanities workflow management toolshave been used in the context of digital text produc-tion For example there are workflows in the PublicKnowledge Project for its Open Journal Systems andits Open Monograph Systems However they arecurrently represented as a linear series of stepsrather than a visual workflow (Fig 2)

There are different approaches to the construc-tion refinement and comparison of workflowsSome methodologies offer better expressibilityusing graph-based models others focus on the

L Frizzera et al

616 Literary and Linguistic Computing Vol 28 No 4 2013

at University of V



ownloaded from

complexity of model checking (Lu and Sadiq 2007)We are interested in producing a visual interfacethat combines workflow management with an ana-lytics tool that could be used by scholars

Based on our previous work we began to con-sider a visual workflow management system as akind of structured surface (Radzikowska et al2011) A structured surface generalizes the familiarconcept of a map with pins The map provides alayer of information and the pins provide anotherlayer which might be used for simple purposes suchas finding restaurants in a city or for more complexones such as making a visual argument or providingevidence for an argument made elsewhere If themap layer is replaced with any other form of datavisualization we refer to it as a structured surfaceIn a workflow manager the structured surface rep-resents the flowchart while the documents areshown as pins Once the pins are combined withthe structured surface the conditions are met foreven further analysis or data visualizations

The resulting interfaces seek to retain the modular-ity and flexibility associated with workflow systemswhile offering a rich-prospect visualization of the col-lection being managed (Ruecker et al 2011) We pro-posed that visual workflows designed as structuredsurfaces could offer an easy tool to track informationduring the processes of production and publicationproviding means for people to gain insight into theirmaterial while also supporting them with some waysof formulating an argument about the data

2 Editorial Workflow

In the pilot stage of this project we used a genericworkflow based on the processes involved in journaleditorship We did not work with any real con-tentmdashjust placeholders and our model of the work-flow was developed based solely on our ownexperiences in publishing in journals Articulatingworkflows is also a non-trivial task and journal edi-tors are typically quite busy The goal was thereforeto get a prototype started so that it could be used ata later stage to elicit iterative improvements fromworking editors

21 TechnologyWe chose to use web standards technologies such asHTML CSS and JavaScript We begin with D3jsan open source JavaScript library for creating datavisualizations The goal was to launch the workflowfrom JiTR a collection management prototype in-tended to support experiments of this kind(Organisciak et al 2009)

22 DesignWe began with the conventional understanding of aworkflow as a kind of flowchart with the surface fororganizing the status of documents showing a pipe-and-flow diagram (see Fig 3) The process is readfrom the left side where each document must beaccepted before it begins traversing the workflow

Fig 1 Algorithmic improvement of workflow layout (from left to right) From Albrecht et al 2010 p 176

Visual workflow interfaces for editorial processes

Literary and Linguistic Computing Vol 28 No 4 2013 617

at University of V



ownloaded from

Each stage of the process represented by a rectangleis an activity such as a decision point in-progresswork or even a standby moment The colours re-flect the status inside the workflow work in progress(dark grey) needs revision (yellow) rejected (red)accepted (green)

The articles are represented by tokens in theshape of bubbles The tokenrsquos position withinthe workflow indicates where along the processthe article is located the size of the bubble roughlysuggests length of the document in words Movingthe mouse pointer over the token shows detailed

information about the article title abstract nameand contact information of the reviewer and thedate when the article went to the reviewer

23 Development resultsWhat we found with this experiment was that toomuch information is contained in the structuredsurface The result is a display that is fine forsmall workflows but can quickly become clutteredand difficult both to design and to subsequentlyunderstand and use As an alternative it is possibleto move some of the information out of the

Fig 2 This screenshot shows a typical workflow in the Open Journal Systems interface

L Frizzera et al


at University of V



ownloaded from

underling structured surface and into the pins Inour next iteration we therefore explored placingmore information in the tokens which allowed usto de-clutter our surface

In terms of technology we also identified a limi-tation Whereas D3js is a good option to createvisualizations using web standards it was not agood fit to create a fully interactive environmentfor our experiment The main limitation has to dowith line breaks As described on the W3C recom-mendation website an lsquoSVG (Scalable VectorGraphic) performs no automatic line breaking orword wrappingrsquo (W3C 2011) Since D3js usesSVG to render graphics on the screen and wewant to show multiple lines of information wedecided to look for other options

3 Orlando Workflow

We hoped for this iteration to produce an experi-ment with more meaningful data so we partneredwith Susan Brown and The Orlando Project One

advantage of this choice was that Orlando had re-cently been going through the process of formallydefining its various production workflows usingUnified Modeling Language (UML) (see Fig 4)Although these are at a level of abstraction thatstill requires translation into a particular instancetheir existence did mean that the Orlando teammembers had been giving recent thought to theirown workflows These had in any case previouslybeen quite well articulated as they were used fre-quently and also formed part of the annual trainingof new Orlando research assistants In discussing thepossibilities we were also able to show the pilotprototype to the Orlando team members whichprovided them with a sense of the direction wewere heading In the end the structured checklistwe used was provided to us by the Orlando TextbaseManager Mariana Paredes-Olea Another advantageof Orlando is that their workflows are predicated onwell-defined roles within the project so that it isonly possible for some people for example tomove a token to the next stage In terms of datafor the tokens we used some of the previous

Fig 3 The first version of the structured surface editorial workflow used D3js to build the interface



at University of V



ownloaded from

Orlando document production jobs which involvedthe writing of XML-encoded original material forOrlando biocritical entries on women writersFifty-four biographical entries were used as asample for the purposes of this prototype

31 DesignSince we moved from a hypothetical workflow to aworkflow predefined by the Orlando Project we hadto redesign both the process and the structured sur-face The sample workflow supplied by Orlando hassome interesting and useful features For examplesome stages are non-sequential but are required atsome point before the document is completed Toillustrate this characteristic we grouped these stepsin clusters with a light grey background Anotherfeature is the existence of optional stages whichcan occur independently of the documentrsquos stageThese are represented in the top of the screen dis-connected from the main stream (see Fig 5) Onefeature we did not explicitly address is the restric-tion of certain actions to particular roles this affor-dance would most likely be provided in futureiterations through the development of rules thatgovern constraints on the system

The workflow loads the collection of documentsand retrieves the metadata Based on this informa-tion the documents represented by tokens are pos-itioned in the workflow To save space the positionof tokens inside a stage is random so they can beplaced one on the top of the other They are colour-coded to show status information Starting a stage(White) Work in progress (Blue) Incomplete

(Red) Completed (Green) A counter box showsthe number of documents held by each stage

As in the first prototype a balloon showing thetitle of the document pops up when the user clickson it However we believed that this informationwould not be sufficient for project managers So wedecided to make available more details about eachdocument A double click makes the token biggerrevealing a circular control panel and general infor-mation such as title collection and current stageand status of the document

The control panel gives three options to the user

(1) Access the History Log which shows thedocument progress in the workflow Thelogs are displayed in reverse chronologicalorder showing when someone performed amodification and providing the name of thatperson

(2) Access the Status Switcher enabling the userto change the status information of thedocument

(3) Tag Mark which turns on a star icon makingit easier to follow a particular token throughthe workflow

In this version the user can also drag the tokento any stage in the workflow regardless of the se-quence By doing that the user automaticallychanges both the stage and the status of the docu-ment Every modification performed in the work-flow is auto-saved and generates a new log entry inthe history The system also prevents tokens frombeing released outside the stages by automatically

Fig 4 The Orlando Workflow for research writing and tagging shown as a UML diagram

L Frizzera et al


at University of V



ownloaded from

moving them back to their former position Thisconstraint means that every token needs to be ac-counted for within the workflow hopefully fore-stalling disruption and confusion

32 Development resultsSome issues arose during the process of develop-ment of this version The most important washow to deal with the high density of tokens insidea stage Whereas it was easy to access one documentin a lightly populated stage in those with a highdensity the fact that the tokens overlap one anotherchallenge the user to find or access the information

Another issue is that the distribution of stages ina linear sequence seems to be a poor representationof all but the most simple workflows In digital textproduction it is common to have parallel actionsoccurring at the same time That is more than oneperson may be working on different stages of thesame article We needed to accommodate for ex-ample a situation where one person is doing theliterature review at the same time that another

person is digitalizing the data for analysis and athird is writing the methodology section

4 Evaluation

To confirm or refute our assumptions and the inter-facersquos effectiveness a user experience study of thetwo workflow prototypes was done The followingsection provides a preliminary report on data col-lected from phase I of a two-phase study In phase Iwe recruited participants who had experience(either as editors or writers) with the editorial pro-cess in academic settings One reason for this ap-proach is that we are hoping that the strategies wedesign might be of use in the general context of largeinfrastructure projects such as the CanadianWriting Research Collaboratory (CWRC) whereusers will have a good knowledge of their ownwork but may have had little or no prior trainingwith the visual workflow system These participantsengaged with both the Editorial Workflow andOrlando Workflow in phase II which is underway

Fig 5 The Orlando Workflow is the second version of structured surface workflow



at University of V



ownloaded from

we are recruiting participants who are involved inXML tagging with the Orlando project Results ofthe full study will be presented at a workflow panelat JADH2013 in Tokyo

41 ProcedureDuring phase I the prototype was tested with nineparticipants at the University of British ColumbiaParticipants engaged with the two prototypes on aMacBook Pro running the OSX 1074 operatingsystem The users were first shown brief video tu-torials that demonstrated and explained the func-tions of the prototype After viewing the tutorialsthe participants completed a short task list for eachprototype The task list provided a set of brief nar-rative statements followed by imperative sentencesasking the participant to use the interface in a wayresponding to the narrative as shown by the follow-ing sample task list item

SETUP The author has sent the article backand yoursquove reviewed it The article requiresfurther revisionINSTRUCTION Please move the article to theappropriate box

A different set of version-specific narrative state-ments was provided for each version of theprototype

Participants were encouraged to talk aloud abouttheir process as they completed tasks A standardthink-aloud protocol was modified in the followingmanner researchers observed and interjected attimes to invite participants to clarify particularcomments or actions This approach has beendescribed as a side-shadowing interview (Luce-Kapler 1996) All interactions during the sessionswere recorded using the screen capture softwareScreenFlow (Telestream 2013) After completingeach set of tasks participants filled out an onlinesurvey that included Likert-scale questions such asthe following lsquoIn your experience as an editor doyou feel that the option to track and view the workhistory of a document in Orlando Workflow wouldenhance your ability to track a document through tocompletionrsquo Finally participants were invited toimagine and draw their own visual workflow process

using a blank sheet of 85rsquorsquo by 11rsquorsquo printer paper Thestudy took 1 hour to complete

42 Data analysisThe audio-visual recordings were reviewed once infull by the investigators to delimit the shared areasof interest and concern among the participants Onthe basis of these commonalities partial transcriptsof the areas of interest were produced These tran-scripts and the survey data form the basis for anumber of general findings from phase I of thestudy We report these below

43 ResultsThe majority of participants found the structuredsurface workflow a pleasant interface 462 of re-spondents to the online survey answered lsquoYesrsquo and231 lsquoMostlyrsquo to the question lsquoDo you like the col-ours in this interfacersquo To the question lsquoDo you likethe layout of the boxesrsquo 308 of respondents an-swered lsquoYesrsquo and 385 answered lsquoMostlyrsquo To thequestion lsquoIn your experience as an editor do youfeel Workflow would enable you to track the progressof academic articles from submission through pub-licationrsquo 462 of participants answered lsquoYesrsquo and308 answered lsquoMostlyrsquo From this we conclude thatWorkflow provides an effective visualization of theprogress of articles towards publication

However there was dissatisfaction among theparticipant group with the smaller details ofmoving and tracking the details of texts In anopen-ended question asking lsquoWhat progress-track-ing feature did you like least and whyrsquo participantsraised a number of issues including the following

lsquoItrsquos a bit odd that itrsquos being done manuallyWhile the ability to move things manuallywould be incredibly helpful being attachedto a system that automates the workflowbased on feedback from the editorsubmitterwould make more sense to mersquo (Response 3)lsquoThe bubbles do not clearly distinguish whicharticle author they represent from an over-viewrsquo (Response 6)

lsquoI would like more information in the bubblesthemselves without having to double click onthemrsquo (Response 17)

L Frizzera et al


at University of V



ownloaded from

Such problems are difficult to explore in detailusing only survey responses However the side-sha-dowing interview procedure provided much dataThese data less structured than the survey responsescan be effectively sorted and contextualized by draw-ing on Barr et al (2002) who proposed a taxonomyof interface metaphors based on Lakoff and Johnson(1980) Barr et alrsquos taxonomy can form an effectivetheoretical basis for both further commentary on andmobilization of the user experience results

Investigators noted that the participants tended toadopt one of two stances in their engagement withthe prototypes The first was that of a critical viewera lsquoreflectiversquo style of response the second was that ofan active user a lsquodirectiversquo style of response Thesestyles of response are pertinent to two distinct levelsof the Workflow prototypes Commentary in a re-flective style informs the underlying structure andmetaphor of the interface whereas commentary ina directive style informs the moment-to-momentpracticality and functionality of the prototype

Reflective styles of response occurred when a par-ticipant did not actively operate the prototype orseek for functionality but instead held back andprovided rich descriptions of the underlying processthat the interface represents Participant 6342 forexample does just this at [0255] when he remarks

lsquoIt looks extremely simple on the one handand useful on the other because it is visuallythere Like a rather complicated index cardsystem but I think having the cards all laidout in front of me like in a flow chart waymakes things look quite a lot simpler and Ido like organizing these kinds of things Overthe years Irsquove edited a bunch of books and Iedited a journal for a few years as well so Irsquovehad a fair amount of experience with the com-plexity of even getting people to review torespond to return revised manuscripts andall that kind of thingrsquo (Participant 6342)

The directive style of response on the otherhand occurred when a user put him or herself inthe role of an actual user of the prototype At thesemoments users entered into the particular task-listitems or sections of the interface as if they wereactive users of the software They probed whether

the movement and communication functionalitiesmade sense on a logical level and tasks could beachieved as described in the task list An exampleof this style of response arose from participant 2751at [0625]

lsquoSo sometimes I am getting a mouse-oversometimes Irsquom not so yoursquoll notice therersquos anice little pulse when I go over and if a hoverfor a very long time nothing Oh actuallyno I can see that now And actually I think it ishappening there but because the backgroundis so dark I canrsquot actually view it And letrsquossee and the red one Okay yeah So some-times visually the size of the bubble andthe darkness of the background is making itless likely to see the effect So the interestingabout that is as a user I saw that effect initiallythat reaction and then I didnrsquot here so Ithought I was stuckrsquo (Participant 2751)

Barr et alrsquos model provides a means to distin-guish what features of the prototypersquos set of meta-phoric meanings the two different styles ofparticipant response are addressing The reflectiveusers contributed information related to what Barret al (2002) defines as the structural metaphor ofthe interface Following Barr et alrsquos (2002) modelwe can interpret such responses as implicit compari-sons the users are making on the basis of similaritybetween preexisting models for managing a work-flow (such as other programs like Open JournalSystems or paper-based workflows) and the appar-ent functionality of the prototypes themselves

On the other hand the active users providedfeedback on what Barr et al (2002) defines as pro-cess and element metaphors

lsquoA process metaphor is used to explain someaspect of system functionality works [sic] Anelement metaphor is a perceivable aspect ofthe user interface that is designed to aid theuser in understanding what process meta-phors are applicablersquo (p 28)

Participant 2751rsquos remarks above on the bubblersquosbehavior can be accounted for as engagement withan element metaphor (the pulse) that was meant tocue a userrsquos understanding of the process of clicking



at University of V



ownloaded from

on and finding greater information about a giventext (represented as a bubble)

Within such a framework we can draw add-itional general conclusions about the workflowprototypes The reflective styles of response tendedtowards acceptance of and a favorable attitude to-wards the underlying structural metaphor of theworkflow prototype There were many mentions ofattributes like lsquosimplersquo or lsquoa good overview of theprocessrsquo However much of the commentariesaround the process and element metaphors wereless positive The process and element metaphorsare an area of concern moving forward with newversions of the prototype

5 Conclusion and FutureDirections

In summary we have been using the concept ofstructured surfaces to investigate both visual andinteractive methods to make workflows more at-tractive flexible and useful for scholars The workhas resulted so far in a series of sketches and twoprototypes of different workflow interfaces Thepurpose is to offer an easy way to track informationduring the processes of document production andpublication while at the same time providing meansto gain insight about the data

For the next versions we are planning to addressthe issues that arose in the user tests and also to im-plement additional new features to enhance the userexperience and the awareness of the system We chosethese features through a combination of user experi-ence results reflecting on the appropriate literaturediscussing the system with our collaborators andexperiencing the shortcomings of the first two iter-ations of the prototype The enhancements are

a Zooming User Interface (ZUI) connection to text analysis tools dependency control of workflow stages tangible interactions and affordances for customization and collaboration

One serious problem we identified is the highdensity of tokens within a stage There are severalpossible ways to address this problem including

reliance on a larger display resizing stages automat-ically to reduce token overlap and implementing aZUI Mandating a large display size is the simplestsolution from the developer point of view but theleast useful from the perspective of the userAutomatic resizing of components would require amuch more robust automated layout mechanismWe therefore settled on the choice of using somesort of semantic zoom or ZUI A ZUI is a form ofintuitive interaction in which varying degrees ofdetail are displayed on screen depending on thezoom level defined by the user ZUIs are commonin map applications such as Google Maps and theyhave started to be implemented as navigation func-tionality in mobile operating systems as well as inMicrosoft Windows 8 With a ZUI the user couldzoom in on a particular densely populated stage ofthe workflow to see more clearly Zooming wouldalso make room for other functionality such as fil-ters and sort options

On the analytics side we are planning to connectthe workflow with Voyant a collection of text ana-lysis tools We feel that having analytics ready tohand could introduce some important changes foreditors making tasks that are currently not evenwithin consideration a simple matter to carry outAn editor might decide for example to take a lookat a KWIC concordance for a term in every acceptedarticle in the current edition or run a wordcloud on all the articles that have been accepted orrejected (see Fig 6) The word cloud of acceptancesmight provide some guidance in writing an intro-ductory editorial for instance while the cloudof rejections might help the editor see whetherthere is a possible gap in reviewer expertise atwork Support for these sorts of curiosity-drivenactivities within the editorial process itself has thepotential to change the nature of the task in a var-iety of ways

Comments and questions from collaboratorsalso suggested some future directions for this pro-ject One of the basic features in workflows is thedependency between stages For example the pub-lisher might intend that an article should automat-ically be published after a final peer-review phase wasdone and this could be accomplished by a rule thatguides the workflow without manual intervention

L Frizzera et al


at University of V



ownloaded from

Without rules to govern the workflow the system isless useful as a guideline and could even potentiallyconfuse the users for instance by implying that someadditional task is required after final clearance byreviewers and before publication With the correctrules in place this automatic adjustment can be pos-sible Similarly the system might prevent the tokensfrom being reallocated to stages that have dependen-cies if the prerequisite stage is not done (see Fig 7)

Customization and automatic layouts could helpusers to produce and organize their own workflows(see Fig 8) In the current version the workflow isprebuilt in the system However ultimately we wantto let the users create workflows for dealing with theirown data A supervising professor for instancecould create a workflow to manage papers confer-ence presentations and student thesis proposals intheir different stages Thus the stages status andmetadata could vary according to the use of theworkflow For this reason automatic layouts maybe useful to enhance the readability by organizingthe stages and reducing crossover in the pathways

Shared workflows could improve the collabora-tive work in digital humanities By sharing

workflows partners and collaborators could followthe jobsrsquo progress and get a sense of how the projectsare moving (see Fig 9) However sharing workflowsposes questions about authority to move tokensaround the workflow as well as the informationsynchronization

Finally we also intend to extend the experimentto different devicesmdashin particular multi-touch plat-forms such as smartphones tablets tables andwalls On one hand the mobility of some of thesedevices could help project managers to keep up onprojects and take actions whenever necessary (seeFig 10) On the other hand big displays onmulti-touch tables and walls may provide a betterway to see and manipulate the documents

At this point we have some experience with twoprototypes and quite a few ideas for an improvedthird However one issue where we have not yetmade much headway is the one perennially facedby software developers and also by researcherscreating experimental systems in the digital huma-nities The issue is the socio-cultural one of technol-ogy acceptance and the diffusion of innovationScholarly writers and academic publishers are not

Fig 6 Connection between Voyant and workflows could help users identify issues during the editorial process



at University of V



ownloaded from

Fig 7 Stage dependency is a fundamental feature that involves rules to control information through the workflow

Fig 8 Sketch showing how to deal with workflow authoring and customization

L Frizzera et al


at University of V



ownloaded from

Fig 10 Multi-touch devices could help project managers to keep workflows updated

Fig 9 The sharing feature could improve collaborative work by indicating work progress



at University of V



ownloaded from

necessarily early adopters of technology and even ifwe were able to produce a system that is generallyapplicable and easy to use it is not likely that thiswill be enough to gain widespread acceptance andadoption Of particular significance in this respect isthe collaborative work with stakeholders in thecommunity which needs to be broadened and dee-pened Revisiting our decisions concerning the coreprocess and element metaphors as suggested by theuser study will also be an essential aspect of the nextphase of the project

AcknowledgmentsThis work was supported by a Major CollaborativeResearch Initiative (MCRI) grant for theImplementing New Knowledge Environments(INKE) project received from the Social Sciencesand Humanities Research Council of Canada(SSHRC) and led by Ray Siemens at the Universityof Victoria Additional support was received fromthe University of Alberta University of BritishColumbia the IIT Institute of Design and MountRoyal University

ReferencesAalst W M P (2004) Business process management

demystified a tutorial on models systems and stand-

ards for workflow management In Desel J Reisig W

and Rozenberg G (eds) Lectures on Concurrency and

Petri Nets Berlin Springer-Verlag pp 1ndash65

Albrecht B Effinger P Held M and Kaufmann M

(2010) An automatic layout algorithm for BPEL pro-

cesses Proceedings of the 5th international symposium on

Software visualization Presented at the SOFTVIS rsquo10

New York USA ACM pp 173ndash82

Barr P Biddle R and Noble J (2002) A taxonomy ofuser-interface metaphors Proceedings of the SIGCHI-

NZ Symposium on Computer-Human Interaction

CHINZ rsquo02 New York USA ACM pp 25ndash30

Berrick S Leptoukh G Farley J and Rui H (2009)

Giovanni a web service workflow-based data

visualization and analysis system [serial online] IEEETransactions On Geoscience amp Remote Sensing 47(1)106ndash13

Bresciani S and Eppler MJ (2009) The risks ofvisualization a classification of disadvantages associatedwith graphic representations of information InSchulz PJ Hartung U and Keller S (eds) Identitatund Vielfalt der Kommunikations-wissenschaftKonstanz Germany UVK Verlagsgesellschaft mbHpp 165ndash78

Lakoff G and Johnson M (1980) Metaphors We Live ByThe Production of Reality Essays and Readings on SocialInteraction University of Chicago Press pp 103ndash14

Lu R and Sadiq S (2007) A survey of comparativebusiness process modeling approaches BusinessInformation Systems 82ndash94

Luce-Kapler R (2008) The sideshadow interview illu-minating process International Journal of QualitativeMethods 5(1) 1ndash16

Organisciak P Geoffrey R Stan R and Susan B(2009) lsquolsquoMashing Texts Supporting collections-leveltext analysisrsquorsquo Paper Presented at the DigitalHumanities 2009 Conference University of MarylandJune 22ndash25

Radzikowska M Ruecker S Brown S OrganisciakP and the INKE Research Group (2011) lsquolsquoStructuredSurfaces for JiTRrsquorsquo Paper Presented at the DigitalHumanities 2011 Conference Stanford June 19ndash21

Ruecker S Radzikowska M and Sinclair S (2011)Visual interface design for digital cultural heritage aguide to rich-prospect browsing Farnham SurreyEngland Burlington VT Ashgate

Telestream (2013) httpwwwtelestreamnetscreenflow

Versteeg R J Alexander N R and Trevor R (2006)Web-accessible scientific workflow system for perform-ance monitoring Environmental Science amp Technology40(8) 2692ndash8

Virine L and McVean J (2004) Visual modeling of busi-ness problems workflow and patterns In Proceedings ofthe 36th Conference on Winter Simulation WashingtonDC Winter Simulation Conference pp 1839ndash46

W3C (2011) Text - SVG 11 (Second Edition) WorldWide Web Consortium (W3C) Retrieved fromhttpwwww3orgTRSVGtexthtml (accessed 20January 2012)

L Frizzera et al


at University of V



ownloaded from

be intermittent tasks with irregular intervals be-

tween work sessions a considerable amount of

time can be spent just in getting back up to speed

on where the work left off Second is the increase inconfidence not necessarily about the status of a

document but about its location among the various

people involved This is particularly important in

cases where the document may be in the hands of

several people simultaneously and it is important to

track how long it is taking in each location so that

bottlenecks can be prevented and the amalgamation

of work carried out efficiently Finally related toreduced confusion increased confidence improved

productivity and other factors there are possible

benefits to the mental and emotional well-being of

the people involvedIn terms of the visual component it is generally

understood that visual representations of data andprocesses can have the advantages of providingan overview of the material insights into theunderlying structure and affordances and benefitsfor non-expert or intermittent users (eg Rueckeret al 2011) However it is also important to avoidthe many potential disadvantages of visualizationBresciani and Eppler (2009) provide a preliminarytaxonomy of gt50 potential cognitive emotionaland social disadvantages which in each case mightbe designer-induced or user-induced

There are several ways in which such improve-ments might be subject to measurement includingcomparisons between using a visual workflow andnot using one for the following

time to create documents publisher hours spent per document published cognitive and affective reports from writers and

editors

This project falls within the larger program ofresearch set out by the Interface Design (ID) teamof Implementing New Knowledge Environments(INKE) and in particular INKErsquos recent work onsupporting the creation and editing of digital ma-terials as opposed to their access and use We havebeen exploring the concept of structured surfaces(Radzikowska et al 2011) to investigate bothvisual and interactive methods to make workflowsmore attractive flexible and useful for scholars We

provide a summary of interactive visual workflowsand how they have been applied and explain whatstructured surfaces are and how the concept can beapplied to workflows

11 WorkflowsMany projects in the digital humanities involveeither digitization or enrichment of existing digitalmaterials The process is usually similar to editorialactivity and can be understood as an ordered check-list or sequence of events or perhaps more accur-ately an ordered graph of events as some can occurin parallel or only after a decision point A re-searcher first enters the raw text then encodesproofs and tests the results There is a process orworkflow to be followed

Visual tools for workflow management thereforeoften also take the form of what is essentially a struc-tured checklist composed of a collection of stageswith transitions including occasional decisionpoints between them In effect they are flowchartsThese tools are widely used in science and business todescribe production process models research in thisarea began as early as the 1970s (eg Aalst 2004)Contemporary examples range from managingearth sciences data at NASA (Berrick et al 2008) toenvironmental performance monitoring (Versteeget al 2006) to modelling business problems(Virine and McVean 2004) There have howeverbeen criticisms both about the lack of visual stand-ards and the poor visual decision choices that resultlsquoOften the construction of a layout for a possiblycomplex workflow is left to the user or the result isvisually unsatisfyingrsquo (Albrecht 2010) The solutionAlbrecht proposes involves better algorithms forautomatic layout (Fig 1)

In the humanities workflow management toolshave been used in the context of digital text produc-tion For example there are workflows in the PublicKnowledge Project for its Open Journal Systems andits Open Monograph Systems However they arecurrently represented as a linear series of stepsrather than a visual workflow (Fig 2)

There are different approaches to the construc-tion refinement and comparison of workflowsSome methodologies offer better expressibilityusing graph-based models others focus on the

L Frizzera et al


at University of V



ownloaded from











at University of V



ownloaded from






L Frizzera et al


at University of V



ownloaded from



3 Orlando Workflow






at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from





4 Evaluation





at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from











at University of V



ownloaded from






L Frizzera et al


at University of V



ownloaded from



3 Orlando Workflow






at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from





4 Evaluation





at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from






L Frizzera et al


at University of V



ownloaded from



3 Orlando Workflow






at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from





4 Evaluation





at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from



3 Orlando Workflow






at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from





4 Evaluation





at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from





4 Evaluation





at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from





4 Evaluation





at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from












L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from














at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from











L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from










at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from



L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from





at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from






























L Frizzera et al


at University of V



ownloaded from