Ray Luo

Chetan Nandakumar

Barath Raghavan

Eugene Ozhinsky

Red Ink – Final Prototype

 

Problem

 

Annotation and revision are integral parts of the writing process. Authors of documents almost always need to send drafts to others for comments, assessment, and editing. The key to success in all of these tasks is collaboration. Authors need an intuitive, timesaving means for annotation, revision, and collaboration on electronic documents.  The current process involves authors creating documents electronically, printing them to pass on to reviewers, and then receiving a jumble of marked up printouts from these reviewers.  Authors then must first decide which revisions to accept and then must manually fold in each change into the electronic original. This process, especially when multiple reviewers are involved, can quickly become tedious and error-prone.

 

Solution Overview

 

The Red Ink document system aims to resolve the above problems by means of a unique pen-based interface. This interface provides a more natural means for annotation, revision, and collaboration on electronic documents.  Red Ink is designed for use by both authors and their reviewers, and provides a single environment in which these users can transparently exchange information.  Reviewers can provide feedback directly on the electronic document using pre-defined pen gestures.  Such feedback will then be available to the author, who can fuse desired changes into the final document using a similar set of gestures.  The entire system is implemented in Java, thereby allowing a degree of portability not available with other possible underlying implementations.  Furthermore, the system is designed with tablet PCs in mind, which will likely be in widespread use in coming years.

 

Tasks

 

Easy Task

To utilize Red Ink, users must choose a document for others to revise. This process involves logging into the system, creating a new account if necessary, going to the proper tab, importing a document to be revised, and adding reviewers to receive the document.

The easy task was chosen to ensure that users are comfortable with basic file management UIs. During user testing and heuristic evaluation, the easy task was a pre-requisite for the more advanced features. It also demonstrated the basic idea of our application: to allow authors and reviewers to transparently exchange information.

 

Moderate Task

As a reviewer, users can revise and annotate documents by using appropriate gestures. This process involves opening up a document in the proper tab, reading it carefully, editing it using the “insert”, “strikeout”, “select”, and other gestures, commenting on it using the comment gesture, and typing in the insertions or comments.

The moderate task was an introduction to the gesture-based interface system. It demonstrated the feasibility pen-based input for emulating the experience of editing documents naturally. More importantly, it showed us, for the first time, whether users could understand how gesture-based interaction works. This task demonstrates the recognition-based engine driving the interface.

 

Difficult Task

As the author, the user can incorporate changes made by other reviewers. This process involves opening a document in the proper tab, accepting or rejecting changes made by reviewers using gestures, and adding any additional annotations.

The difficult task was designed to test the user's ability to use advanced gestures that were designed to interact with other pre-made annotations as opposed to the text itself. This task is difficult because users are asked to understand the way the “accept” and “reject” gestures work in their relationship to the seemingly unrelated set of annotations seen on the screen. Thus, the difficult task tests whether our interface is as intuitive as paper and pencil.

 

Design Evolution

 

As expected, the design of the Red Ink system has changed several times through the course of the semester.  The main interface originally was a completely encompassing interface including document and reviewer management, network components such as publishing, and gesture-based document annotation.  Although the structure of gesture recognition has largely remained the same through the semester, there are several elements of the interface that have changed.

 

Changes between contextual inquiry and low-fi prototype

Originally, our interface was designed with one main pane containing all aspects of “collaborative annotation”.

 This monolithic interface allowed annotation of other users’ documents, sending out documents to be revised, and editing documents in general.  Furthermore, this interface never clarified whether annotations would be collaborative in the “real-time” sense.  We decided against a real-time system since revision is not currently a real time process and also because real time interaction did not grant anything significant to the task.  In addition, we realized that a more conventional document management interface was necessary since allowing document management within the context of a pen based interface is somewhat confusing.  Lastly, we also had not decided whether the interface would be a stand-alone application or part of a Web application.  Upon design of our low-fi prototype, we realized that the interface needed to be a stand-alone application, since web applications are quite limited in their functionality. 

These design decisions required us to break up the application into 2 primary sections: the document management section and the gesture based annotation section. A login screen leads the user into the document management window.

 

  

Here the user can view the properties of the different documents, send them to revisers, and other such tasks. The user can also view a document a document, and it will then be loaded in the annotation pane using gesture-based interaction.

 

Changes between low-fi prototype and interactive prototype #1

We found two primary difficulties during the low-fi testing. The first was not related to the design of the interface, but rather it was the lack of feedback of a paper prototype; the second was the unfamiliarity users had with using a gesture-based interface.  Although the low-fi interface presented the user with an experience similar to the process of revising a document on paper, using a pen to make gestures on top of a transparency remained very unnatural.  Due to the lack of useful feedback during the user testing, the gestures and annotation pane interface remained largely unchanged between the low-fi and interactive prototypes.  Since the interactive prototype was stand-alone, we employed Wizard of Oz techniques to fake the behavior of the gesture recognition.

 

We did gain some very useful information regarding the document management interface. This interface contained a large number of buttons, and through this phase of user testing, we found that the large volume of buttons confused the user. To rectify this problem, we provided a much simpler alternative in the first interactive prototype.

 

Changes between interactive prototype #1 and interactive prototype #2

After receiving feedback from our peers on our first interactive prototype, we made several changes to the design.  Lack of system status and error recovery were two of the primary heuristics that our interface violated.  We added dialogs that allow the user to cancel, and in addition, we added document removal confirmation. 

 

In our interactive prototype #1, it was unclear what the document list box was actually displaying.  In this new interactive prototype, we include information about the number of reviewers per document, and the number of new revisions, if any.  This additional information makes it clear to the user that this document is available to be revised or has been sent to them to be revised.  The most significant implementation difference between interactive prototype #1 and interactive prototype #2 was that no Wizard of Oz techniques were used at all.  Despite all the feedback we received, most of the changes were rather minor.

 

 

Changes between interactive prototype #2 and interactive prototype #3

Between these two prototypes, we conducted a pilot usability test to find any other design problems with the Red Ink system. This test was a bit premature since we had not completely fleshed out the interface, but it still provided us with some good feedback. We found that the users still had a bit of difficulty switching into the paper-based user model using gestures for interaction instead of toolbars. To ease the user into the interface, we included a tutorial when she signed up as a new user and also implemented the help feature in the annotation pane.

We also found that the user did not always perform the gestures properly and was confused since not all of the gestures on the left pane were implemented. To clear the confusion, we grayed out the gestures not implemented and placed thick dots at the critical points in the gesture. We found that the user easily got frustrated when she was trying to do something that she thought was correct but the machine was misunderstanding. To solve this problem, we added a status window displaying the interpretations of the machine. This window also allowed us to display the content of comments and insertions.

In addition, we cleaned up the gesture recognition so that tapping is now ignored and included features that we did not have the time to implement before such as the multi-level undo, the button bar at the bottom, and buttons for future features in the comment window (ie. audio annotation). Lastly, we cleaned up some of the document management panes by re-labeling the “Edit Reviewers” button as “Send to Reviewers”, making the corresponding pane more intuitive, converting the add document dialog into a more standard form, and making the button arrangement more aesthetic.

 

Overall design changes

On the surface, it may be surprising to note that the design ideas have not changed drastically since the initial sketches.  The primary annotation interface, which allows gesture based document editing, still relies upon almost the same gestures as we originally envisioned.  Although some gestures have had to been changed due to technology limitations, most gestures look almost the same as they did in the original hand sketches. The interface now contains the document management system, which allows the user to view and select documents following the usual graphical user interface paradigm. 

 

 

Scenarios

 

Easy Scenario

Alice, a young film critic at the Los Angeles Times, has received an invitation from her long time friend Ray to edit his one page final paper on Spinoza's theory of the nature of the mind and its implications for contemporary film-making. She has edited numerous papers for Ray and has a good knowledge of the common mistakes of Ray's writing. Alice will perform these tasks using a pen on a tablet PC. She logs into the red ink system:

From the main menu, she chooses Ray's document from the list of documents to review.

This action brings up an editing panel consisting of the document and various controls and displays.

Knowing Ray's tendency to switch the first and last paragraphs around, Alice checks the last sentence on the essay and moves it to the front. Next, she goes through Ray's paper, deleting his excessive uses of academic lingo. Specifically, she will delete words such as "sophisticated," "unmitigated," and "repugnance." Given Ray's tendency to mix up his words, Alice will also make sure that the ordering of phrases are correct, for example "very well be made". She will also read over the document a few times, making general comments on Ray's writing, warning him, for example, against the use of too many exclamation marks She will also italicize book and film titles in the essay appropriately on the document. Finally, she will click on "done", which hides the file and notifies Ray the next time he logs on that she has finished editing the document.

 

Moderate Scenario

Ray, a philosophy major at UC Berkeley, has just finished his one page summary-critique of Spinoza's theory of the human mind and its relations to modern cinematic thoery. He is uncertain about his short paper, which accounts for 50% of his Film 101 course. He decides to upload his document to the Red Ink System to be reviewed by his old high school friend Alice, who happens to be a film critic. After getting into the system, Ray opens the existing document to be uploaded.

To make sure Alice catches all the mistakes for him, Ray opens up the annotation pane to make authoring comments on the document.

He circles the words "substance" and "attribute" and asks, in a comment, if their definitions were easy to understand.

He asks the entire quotation from Wild Strawberry and asks whether the material contributes to a general understanding of the essay He also makes other general comments dealing with the structure of the essay and diction. Finally, he exits the program and awaits the response of Alice in a future session.

 

Difficult Scenario

Fred has been notified that Alice has finished editing his essay. He retrieves the revised document and begins filtering through the comments and revisions.

Fred then decides to walk through the comments one by one, judging for their significance. For example, he accepts Alice's comment on the use of excessive exclamation marks in "No way!!!!". But he rejects Alice's italicized version of "Wild Strawberries" because he thinks that film titles should be in quotes, not italicized.

 

Final UI Design

 

Our final UI design consists of 2 distinct sections. The first is the Document Management Pane and the second is the Annotation pane using gesture-based interaction.

 

Document Management Pane

The Document Management Pane is designed to organize and manage all of the documents important to the user in an easy and intuitive way. The documents are divided into two categories – “Document You’ve Written” and “Documents to Revise”, and these two categories are visually separated using a tabbed interface.

 

The look and feel is designed to be familiar to the user by utilizing well-known paradigms in window-based interaction.

The “Documents You’ve Written” section treats the user as the author of documents that she wants feedback on. So there are lists of documents that have been imported by the author herself, and for each document listed, there are a few important properties of the document visible such as the number of reviewers and if there are any new revisions.

This interaction scheme requires that the user first select a document and then use the buttons on the left to act on it. In addition to these buttons acting on the documents in the list, there are others below that are independent of the selection. To delineate the two groups of buttons to the user, we have boxed one group as acting on the selected document and located the other group of buttons underneath.

The user can modify the list of documents by removing selected items, editing the reviewers of a selected item, and adding in a new document. Most importantly, once a reviewer has made revisions, the author can click on the “View/Annotate” button after selecting a document and visually view the annotations made. There are also a few other tools such as a search, help, and exit.

The “Documents To Revise” section treats the user as a reviser; in other words, other people have authored documents and want this user’s feedback on them.

 

This pane is quite analogous to the “Documents You’ve Written” Pane. It contains a list of documents sent to the user from different authors along with buttons laid on the side. In the document list, the important fields of each document are its author and the date received. The user can remove a selected document as well as annotate it. There are also the search and help buttons as on the other tabbed pane.

 

Annotation Pane

            Pressing the View/Annotate button from either one of the tabbed panes brings the user to the annotation pane. This pane is the centerpiece of the interface and quite different from traditional computer-based UI’s as it relies heavily on gesture-based interaction.

If the View/Annotate button is pressed from the “Documents To Revise” Pane, the reviser’s version of the annotation pane is seen. Here, the essay to be revised occupies the center of the screen, a help pane describing the possible gestures occupies the left and the button bar and status window occupy up the bottom of the screen. The interaction scheme aims to be more natural than current solutions, and so the interface is directly modeled after paper-based revising. The user is expected to treat the window as a piece of paper, and so she makes gestures on the document itself.

            These gestures are all interpreted by the computer and feedback on the interpretation of the gesture is immediately returned when the computer redraws the stroke. The user uses the left help pane as a guide to what gestures are recognized by the system, and how to properly draw them; the thick dots on each gesture are designed to give the user a guide as to what critical points the gesture interpreter is looking for. All of the implemented gestures are highlighted in blue and the unimplemented ones are grayed out.

            In brief, the reviser sees that someone has sent her a document when it is listed in the document to revise pane. The user selects the document, clicks on view/annotate and the annotation pane pops up; then the user revises the document in a similar fashion as she would with paper. She makes gestures on the document itself using the help pane on the left as a guide to the gestures, and once she is finished, she clicks on done.

 

Error Handling

            However, gesture recognition is not always perfect and errors do occur on part of the computer and the human. To account for this possibility, we have implemented a “multiple event undo” button on the button bar. This will grant the user the flexibility to remove an annotation or string of annotations previously made. In addition, we have included a status window at the bottom of the screen to give the user a glimpse into the interpretations of the computer.  This feature is necessary so that the behavior of the computer does not seem as mystical and the user has more input as to why an error occurred. We hope that this window will ease user frustration if the system does not operate optimally.

            Lastly, if the user is still confused or might be having trouble with our interface, we have implemented a help feature accessible from the button bar.

           

Gestures

            The gestures we are using are modeled after the gestures we saw being used in the contextual inquiry stage. We wanted to make the gestures in our system similar so that they would be intuitive to the user. However, due to the limitations of the gesture recognition technology, we could not implement the gestures exactly. The technology required all of the gestures to be single strokes as opposed to multiple strokes, and it also required that the gestures be quite distinct from one another.

It is important to note that the insert and comment gestures are a little different than the other gestures. These two gestures require a pop-up window in addition to the gesture itself. So upon insert, for instance, one must place a carrot where an insertion is desired and then a window will pop-up asking for input of the content to be inserted. Right now, text input is all that is implemented, although we have provided the buttons for other desired features such as audio annotation and free hand text. Once this is complete, the user can tap on the insertion icon and the content of the insertion will be displayed in the status window.

Authoring View

            If the view/annotate button is pressed from the “Documents You’ve Written” Pane, the authoring view of the annotation pane is viewed. This view is almost identical to the reviewer’s view; the main difference is in the task being carried out. In the reviewer’s view, the task is to revise the document, whereas in this screen, the task is to accept or reject changes made by other reviewers.

            The author sees all revisions made by reviewers in blue, and each revision is contained in a gray box. The author can place a check or an x centered in the box to respectively accept or reject a given revision. If the revision is accepted, than it is automatically folded in, ie. If the user accepts a strikeout then that piece of text is deleted. If the revision is rejected, then the revision mark is erased from the document.

            Another feature unique to the authoring view is the reviewer’s button in the button bar. This feature is designed to manage revisions made by multiple reviewers on the same document. The author can select which reviewers she wants to see revisions from. This feature is currently unimplemented.

 

 

Other Features

            In addition, we have a few other features that we thought would be useful. One is a save feature so that changes are not lost if the computer crashes in the middle of the task. Another is the search feature that helps the user manage longer documents.

 

Features Not Implemented

Several of our initial design plans were left off of our final prototype due to time and technology constraints.  In our initial design, we had focused our attention upon the actual content annotation, such as being able to mark up images, video, as well as regular text in a rather sophisticated manner.  Firstly, we realized that this would be overwhelming to use for users; secondly, implementing such features would not truly address usability issues, but rather mere time-consuming technology ones.  For this reason, features such as audio annotations, while convenient for the user, are not fundamental in our interface design and thus have been omitted.

Other features we planned have been omitted because they would only serve to confuse the user unless completely implemented.  For example, all of our prototypes, both lo-fi and hi-fi have included a document view slider.  This slider was made to allow switching between summary view and document view.  The intent was that the user could view a summary of all the annotations and revisions made to a document at a glance without going through the document to find all the individual marks made.  However, since in all of our prototypes, the features has been unimplemented, users have had no idea what it does, and for some reason were always compelled to attempt to use it in vain. Furthermore, we found that a slider would not be the right widget for this purpose, since a slider usually implies the use of multiple intermediate modes, whereas our summary/document view is more bi-modal. 

One feature that we have not implemented is managing the revisions of multiple reviewers. Although this feature is quite important to the task and we did demonstrate it in the first interactive prototype, we did not have the time to implement it.

Other superfluous features such as signature recognition for login have been left off since they do not significantly affect the user experience in either a positive or negative way. Also, since we are focusing in on the interactions, we displayed the save button but did not choose to implement it.

Finally, as expected, the system does not implement backend network connectivity and a user database.  This does not affect the interface, but it causes the search and save features to be non-functional, and in addition, the “Documents You’ve Written” tab and “Documents To Revise” tab do not accurately reflect documents as if they had been written by the user or sent to them by a peer.

 

Tools

We used Java 2 Version 1.3 and the Swing Toolkit to implement our interface. The Java programming language is easy to use and fairly productive. The Swing documentation was comprehensive and mostly comprehensible. We used the Forte for Java CE 1.0 development environment to manage our projects. It was easy to learn but difficult to customize. Satin Version 2.2 and Quill were used to make our gesture sets and interpret gestures based on the set we made. Quill had a easy to use and efficient interface. We used Satin only for one specific task of providing us with a ranking of possible gestures. We chose Swing, Satin, and Quill because they allowed us to move from basic gesture sets to a full-fledged interface without having to learn another programming language. Java was the programming language of choice because of its simplicity and universality.

The Swing package is enormous. It is difficult for the ordinary programmer to keep track of its entire set of relevant interface components. Hence we relied on Forte for much of the GUI level development. Unfortunately, Forte is tedious for managing multiple projects. It is also difficult to customize and has a steep learning curve. For example, Forte does not allow us to delete certain code. If we delete code using another application, Forte regenerates it unless we delete the form as well. It also took a long time to simply set up our programming environments because of the rigidity of Forte. Forte has poorly designed efficiency features such as a limited set of customizable short cuts. Satin is large and complicated. Hence we only used what was absolutely necessary to our GUI: Sheet and RubineInterpreter. Some of the arguments to functions in Satin are not very intuitive. Quill was a bit too simple - we had hoped to create multiple-stroke gestures, but quill will not work optimally if we had.