Beelders Using Vision And Voice To Create A Multimodal Interface For Microsoft Word 2007

Using Vision and Voice to Create a Multimodal Interface
for Microsoft Word 2007
T.R. Beelders and P.J. Blignaut
The University of the Free State, Bleomfontein, South Africa
{beelderstr; pieterb}@ufs.ac.za

Abstract interface should be customizable and allow users to select any
combination of interaction techniques which suit their needs.
There has recently been a call to move away from the standard The premise of the research study is not to develop a new word
WIMP type of interfaces and give users access to more intuitive processor but rather to incorporate additional interaction tech-
interaction techniques. Therefore, it in order to test the usability niques, besides the keyboard and mouse, into an application
of a multimodal interface in Word 2007, the most popular word which has already been accepted by the user community. This
processor, the additional modalities of eye gaze and speech rec- will allow for the improvement of an already popular product
ognition were added within Word 2007 as interaction tech- and stimulate inclusiveness of non-mainstream users into the
niques. This paper discusses the developed application and the mainstream market. Therefore, one aim is to determine whether
way in which the interaction techniques are included within the it is possible to customize an interface to such an extent that all
well-established environment of Word 2007. The additional user groups are catered for with an all-inclusive interface.
interaction techniques are fully customizable and can be used in
isolation or in combination. Eye gaze can be used with dwell The research study is still in the beginning phase where devel-
time, look and shoot or blinking and speech recognition can be opment of the tool is underway. Therefore, for the purposes of
used for dictation and verbal commands for both formatting this paper, the application as it has been developed will be the
purposes and navigation through a document. Additionally, the main focus. The paper will, however, conclude with a short dis-
look and shoot method can also be combined with a verbal cussion of the next phases of the research study.
command to facilitate a completely hands-free interaction. Mag-
nification of the interface is also provided to improve accuracy 2 Interaction Techniques
and multiple onscreen keyboards are provided to provide hands
free typing capabilities. Using a physical input device in order to communicate or per-
form a task in human-computer dialogue is called an interaction
Keywords: Eye-tracking, speech recognition, usability, word technique [Foley, et al., 1990 as cited in Jacob, 1995]. The inte-
processing, multimodal raction techniques of speech recognition and eye tracking will be
included in a popular word processor interface to create a mul-
1 Introduction timodal interface as a means to determine whether the usability
of this product can be enhanced in this way.
The word processor application has evolved substantially since
its initial inception and since then has undergone a virtual meta- Although this approach has received limited attention thus far,
morphosis to achieve the capabilities that are available in these the multimodal approach has always focused on the develop-
applications today. As an integral part of everyday life for many ment of a third-party application, for example EyeTalk [Hatfield
people it caters for a very diverse group of users, Furthermore, and Jenkins, 1997]. Contrary to this, this study will use an al-
users with disabilities or needs other than those of mainstream ready existing application, namely Microsoft Word ©, which
users are not always taken into consideration during system currently enjoys a high prevalence in the commercial market.
development and often have to compensate by using specially
designed applications which do not necessarily compare with the 3 Development environment
more popular applications. This study therefore aims to investi-
gate various means to increase the usability of a word processor The development environment used was Visual Studio 2008,
for as wide a user group as possible. making use of the .NET Framework 3.5. Visual Studio Tools for
Microsoft Office System 2008 (VSTO) in C# was used for de-
For this reason, the interface of the most popular word processor velopment. VSTO allows programmers to use managed code to
application will be extended into a multimodal interface. This build Office-based solutions in C# and VB.NET [Anderson,
interface should facilitate use of the mainstream product by 2009]. In order to incorporate the speech recognition the Micro-
marginalized users, whilst at the same time enhancing the user soft Speech Application Programming Interface (SAPI) with
experience for novice, intermediate and expert users. Ideally the version 5.1 of the SDK was used. The SDK provides the capa-
bility of compiling customized grammars and accessing the
functionalities of the speech recognizer. In order to provide gaze
Copyright © 2010 by the Association for Computing Machinery, Inc.
Permission to make digital or hard copies of part or all of this work for personal or
interaction Tobii SDK 1.5.4 was used. For magnification pur-
classroom use is granted without fee provided that copies are not made or distributed poses, which will be discussed in an upcoming section, the
for commercial advantage and that copies bear this notice and the full citation on the commercial product Magnifying Glass Pro 1.7 was chosen as a
first page. Copyrights for components of this work owned by others than ACM must be relatively inexpensive solution but primarily based on the fact
honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on
servers, or to redistribute to lists, requires prior specific permission and/or a fee.
that it was one of the few applications which incorporated click-
Request permissions from Permissions Dept, ACM Inc., fax +1 (212) 869-0481 or e-mail able areas within the magnified area which are then correctly
permissions@acm.org.
ETRA 2010, Austin, TX, March 22 – 24, 2010.
© 2010 ACM 978-1-60558-994-7/10/0003 $10.00

173

transferred to the underlying area. This is essential in the devel- continued use of the traditional interaction techniques. As illu-
oped product as the magnification will increase the accuracy of strated in Figure 1, an extra tab was added to the established
cursor positioning via eye gaze and correct interpretation of user Microsoft Word ribbon. This tab (circled in red) was named
intention and requiring the user to disable magnification before Multimodal Add-Ins.
clicking on the interface would negate all the advantages gained
from magnification The new tab provides numerous options to the user to select
which additional interaction techniques they would like to use
The aim of the development process was to incorporate speech (Figure 1). As is evident from Figure 1, complete customization
recognition and eye tracking as additional interaction techniques of the techniques is allowed via selection of any combination of
in the Microsoft Word environment. The user should also be techniques as well as in what capacity the techniques must be
given the freedom to determine in which combination the inte- implemented.
raction techniques must be used, while still having the option of

Figure 1: Multimodal Add-ins for Word 2007

Additional tools which are available to enhance the user expe- Speech recognition
rience are a magnification tool and an onscreen keyboard which
can be displayed at the bottom of the Word document. The mag- The user has the option of enabling the speech engine so that
nification tool magnifies the immediate area under the mouse Microsoft Word can respond to verbal utterances. In terms of the
cursor, thereby providing increased accuracy for users with customizable options, the user can toggle between dictation
weak eyesight and those making use of the gaze sensitive inter- mode and command mode. In dictation mode, the speech recog-
face. Magnification is available when using the mouse or when nition is implemented in the well-known method of capturing
using eye gaze as an interaction technique. The use of the mag- vocalizations, translating those vocalizations into text and writ-
nification tool is entirely at the discretion of the user who is ing the result to the currently activated document in Microsoft
capable of turning magnification on and off at will or as needed. Word. In order for the dictation mode to be effective the user
Magnification can be toggled using either a verbal command or must select a previously trained profile. A unique profile can be
the available button on the ribbon. trained through the Windows Speech wizard. All the available
speech profiles are provided in a drop-down box on the multi-
Onscreen keyboards are available as an alternative to using a modal add-in tab for the convenience of the user.
traditional keyboard. The onscreen keyboard can be used either
through use of the traditional mouse or to achieve hands-free In command mode, a grammar is activated which accepts only
typing using eye gaze or a combination of eye gaze and speech isolated commands and responds to these in a pre-determined
recognition. The final adapted interface, as envisioned in use manner. Command mode provides the functions of cursor con-
when the on screen keyboard is in use, is shown in Figure 2. trol, formatting capabilities and certain document handling ca-
pabilities. Several different commands are provided which have
The layout of the onscreen keyboard can be changed to either a the same application reaction, thereby contributing to further
traditional QWERTY keyboard layout or to an alphabetic layout. customization for the user as they can determine which the most
Each keyboard contains all 26 alphabetic letters, a Space bar, desirable command is for them to use. Moreover, simple cursor
Backspace and Delete keys as well as special keys which simpli- control is provided by providing directional commands but more
fy movement through the document. Special keys which are complex cursor control is also provided by allowing line selec-
provided are Page up, Page down, Home and End. The user can tion and movement of the cursor as though control keys (such as
also toggle between upper case and lower case by activating and Shift) are being pressed in combination with the verbal com-
deactivating the CAPS lock key. A Select All key is provided as mand. These types of commands will simplify selection of text
a means for the user to select all the text in the document. The and provide verbal commands for complex key combinations
two red arrows in the lower left corner of the keyboard (Figure which are not always known to novice and intermediate users.
2) change the size of all keyboard keys in decrements and in- For example, the word “Bold” causes the activation or deactiva-
crements of 10 pixels respectively, thereby providing even more tion of the bold formatting style. Similarly the words “Italic” and
customization of the keyboard for the user. Auditory feedback in “Underline” activate or deactivate their formatting style. Words
the form of a soft beep is given when a keyboard key is clicked such as “Cut”, “Copy” and “Paste” allow for text manipulation
on. and are their subsequent actions are of course the cutting or co-
pying of the currently selected text and the pasting of the clip-

174

board contents at the position of the cursor. More complex current cursor position. Cursor control is achieved through the
commands for text selection are available such as “Select line”, commands “Left”, “Right”, “Up” and “Down”. Verbal com-
which selects the whole line on which the cursor is situated, mands can be issued in sequence to perform relatively complex
“Select word”, which selects the word nearest to the right of the document manipulation.

Figure 2: Adapted interface of Word 2007 when the onscreen keyboard is activated

Eye-tracking mouse click is simulated. When selecting the look and shoot
method, the user can position the cursor using eye gaze and then
The eye tracker can be calibrated for use directly through the press the Enter key to simulate a left mouse click. This would
Microsoft Word interface. This increases the usability of the have the effect of either placing the cursor at the position of the
application as the user is not required to move between applica- eye gaze or clicking on the icon directly under the eye gaze of
tions to achieve their goal of using gaze as an interaction tech- the user. The third option available to the user is that of blinking.
nique. Since the word processor is the focus of this study, this In this scenario, the user fixates on the desired object or position
meets the requirement of the research question scope. The user and then blinks their eyes to simulate a left mouse click.
has the option to activate eye gaze which can then be used to
position the cursor in the document or over an object to be ma- Multiple interaction techniques
nipulated. Customization is provided by allowing the user to
choose the activation method. For use purely as a single interac- When the user selects No activation (Figure 1) via eye gaze that
tion method, the choices of dwell time, look and shoot and implies that they will instead be using voice commands to re-
blinking are provided. When dwell time is selected, the user is spond to the current eye gaze position of the user. In this in-
able to set the interval of the dwell time (see the Sensitivity Set- stance, the speech recognition must also be enabled and the user
ting text box o Figure 1). This provides additional customization can then issue verbal commands to move the cursor to the cur-
as the user can determine the speed with which they are most rent gaze position which is analogous to executing a left mouse
comfortable and leaves the option for adjusting this interval as click at that position. In this way, it is possible for the user to
the user gains more confidence and experience with gaze based place the cursor at any position in the document, or to click one
interaction. The interval can be changed at any time during the of the Microsoft Word icons on the ribbon. The verbal com-
use of the application. Dwell time requires the user to fixate on a mands of “Go”, “Click” or “Select” all simulate a left mouse
position for the set time of the dwell time interval before a left click at the button closest to the current gaze position. In this

175

way, the user is free to choose the command which they find Furthermore, to further investigate the usability of the newly
most suitable for them. developed application, user efficiency effectiveness can be
measured in a within-subjects experiment by requiring users to
In most instances it is envisioned that the onscreen keyboard complete identical tasks in both the commercial Microsoft Word
will also be activated under these circumstances. When the on- and the new multimodal Microsoft Word.
screen keyboard is activated in conjunction with the eye gaze,
visual feedback is given to the user to indicate which button will Moreover, the usability of the various interaction techniques will
be clicked when the verbal command is issued. With each fixa- also be analyzed to determine which combination of the interac-
tion that is detected within the boundaries of the keyboard, the tion techniques provides the most usable interface – if any. User
button which is closest to that fixation is determined to be the satisfaction will be measured through means of a questionnaire
target and a shape is displayed in the centre of the button. The in order to gauge user reaction, both in a short-term and long-
user can also select which shape they would like to use for visu- term exposure period.
al feedback. The available shapes are a solid square, a hollow
square, a solid disk and a hollow circle. The hollow shapes do 5 Summary
not obscure the letter of the key and in so doing provide the
necessary visual feedback whilst still allowing the user to see the
letter which will be written to the document. Feedback is only A multimodal interface was developed for Microsoft Word in
order to eventually determine whether the usability of this appli-
given on the keyboard to minimize interference during normal
cation can be enhanced for mainstream users whilst simulta-
document browsing. In order to achieve increased stabilization
of the feedback within a targeted object, the algorithm as sug- neously providing an adaptable and usable interface for disabled
users. For these purposes, eye tracking and speech recognition
gested by Kumar (2007) was used.
capabilities were built into the Word interface. These interaction
If the user is satisfied that the correct button has been deter- techniques can be used in isolation or in combination and the
way in which they are used can be customized in a number of
mined they can then issue any of the verbal commands to simu-
ways. Once the development has been completed and measure-
late a left mouse click. The letter shown on the keyboard is then
written to the document at the current cursor position. ments can be captured automatically in the background during
user interaction, a longitudinal usability study will be undertaken
Both disabled and able-bodied users will be included in the
4 Where to next?
sample and will be required to complete a number of practice
sessions with the application over a prolonged period of time.
As previously mentioned, the research study is still in the pre-
After each session, participants will be required to complete a
liminary stages of an empirical study. An application has been
developed to investigate the effect of multimodal interaction number of tasks, during which measurements will be captured
for further analysis In this way, it will be possible to determine
techniques on the usability of a mainstream word processor ap-
whether users are able to improve their performance on the sys-
plication. Further enhancements to the application will include
the expansion of the keyboards to include numerical keys and tem over an extended period – in other words, whether the sys-
tem is usable. Additionally, user performance between the new
the magnification will be refined to respond to eye gaze and
application and the commercially available application will be
voice commands. More voice commands will be provided for,
particularly for commands that currently have shortcut keys compared to determine whether they can achieve comparable
performance on both the systems. In this way, it will be possible
assigned to them, such as Save and displaying certain dialog
to determine whether a popular commercial application can be
boxes.
fully exteneded into a worthwhile multimodal application which
caters for a diverse group of users comprised of both disabled
Additionally, a back-end will be written for the application
and able-bodied users.
which will capture certain measurements which can be used for
usability analysis. Measurements such as the number of errors
made during a task, the number of actions required and the per- References
centage of the task completed correctly will automatically be
saved to a database for further analysis. ANDERSON, T. (2009). Pro Office 2007 development with VSTO.
APress: United States of America.
Once the application has been completed, user testing will
commence. Both disabled and non-disabled users of a local uni- HATFIELD, F. AND JENKINS, E.A. (1997). An interface integrating
versity will be approached to participate in the study. A longitu- eye gaze and voice recognition for hands-free computer access.
dinal study will be conducted whereby the participants will be In Proceedings of the CSUN 1997 Conference.
required to spend periods interacting with the system. After each
exposure to the system, users will be required to complete a JACOB, R.J. (1995). Eye tracking in advanced interface design.
number of tasks for which measurements will be captured. In In Virtual Environments and Advanced interface Design, W.
this way, the learnability of the study can be measured over a Barfield and T. A. Furness, Eds. Oxford University Press, New
period of time by comparing the results of these sessions to de- York, NY, 258-288.
termine if user performance increases in correlation to user ex-
posure to the application. Since it is expected that there will be a KUMAR, M. (2007). Gaze-enhanced user interface design. PhD
learning curve associated with the application, it is deemed more Thesis, Stanford University.
applicable to capture usability measurements over a period of
time rather than only after a single session with the application.
In order to determine whether the application succeeds in pro-
viding for disabled users whilst simultaneously providing for a
better user experience for mainstream users, it is imperative that
users from both these demographics be included in the sample.

176

Beelders Using Vision And Voice To Create A Multimodal Interface For Microsoft Word 2007

More Related Content

What's hot (10)

Viewers also liked (20)

Similar to Beelders Using Vision And Voice To Create A Multimodal Interface For Microsoft Word 2007 (20)

More from Kalle (20)

Beelders Using Vision And Voice To Create A Multimodal Interface For Microsoft Word 2007