Training people with aphasia to use word prediction software for email writing Lindsey Thiel Karen Sage Paul Conroy

Training people with aphasia

to use compensatory writing

software for email writing

Dr Paul Conroy

Dr Lindsey Thiel

Aphasia and the internet

Recent survey study by Menger, Morris & Salis (2014):

• PWA use the internet less than people with stroke and

no aphasia

• Only 22% of PWA aphasia rated their own internet skills

as good or excellent, compared to 74% of UK

population (OXIS, 2010)

• Main barrier to using the internet in PWA was their

aphasia

• Most popular type of internet use in internet users with

aphasia was email, but 71% needed help with this

Writing therapy studies

Recent review (Thiel, Sage & Conroy, 2015)

found 62 writing therapy studies

– 54 evaluated impairment-based therapies

e.g. Ball, de Riesthal, Breeding, & Mendoza (2011)

Beeson, Higginson & Rising (2013) Tsapkini & Hillis (2013)

– 13 of these trained functional writing skills

e.g. Mortley, Enderby and Petheram (2001) Robson,

Marshall, Chiat, & Pring (2001) Panton & Marshall (2008)

– 3 trained writing for the internet or mobile phones

Beeson et al. (2002), Beeson et al (2013) (Greenwald,

2004)

Writing therapy studies

• 14 evaluated assistive technologies for writing

• Included

– electronic spelling aid (Beeson, Rising, Kim, & Rapcsak, 2010),

– Lightwriter (Jackson-Waite, Robson, & Pring, 2003)

– predictive writing software (Behrns et al., 2009)

– voice recognition software (Estes & Bloom, 2011)

– C-Speak Aphasia (Nicholas et al., 2011)

– speech synthesiser software (Armstrong & Macdonald, 2000)

• 3 trained writing for the internet (Estes & Bloom, 2011;

Nicholas et al., 2005; Nicholas et al., 2011)

Our focus: predictive & editing writing software

Penfriend Ltd.

Predictive writing software

• Designed to assist people with physical disabilities with

writing (e.g. Higginbotham, 1992)

• Also used with children with developmental language or

learning disabilities (e.g. MacArthur 1996)

• Studies with aphasia

– Armstrong & Macdonald, 2000

– Behrns et al., 2009

– Mortley et al., 2001

– Murray & Karcher, 2000

• Changes to written texts such as increased length and

accuracy or richer in terms of content

Motivations for study

• People with aphasia are often excluded from using the

internet due to their aphasia

• Access to internet could mean more opportunities for

communication and better access to information for PWA

• Lack of therapy studies measuring functional outcomes of

writing therapies

• What is the potential of predictive writing software to

improve writing in terms of

spelling accuracy and speed?

Outcome measures

• Functional writing (email) task

• Frequency diary

• Written picture description

• CAT disability questionnaire

• Computer and internet skills

• Keyboard skills

• Self-correction of spelling

• Interviews

Functional writing task

1. Write an email arranging to meet a friend at a

certain time, place and date.

2. Write an email to a friend telling them about a

recent holiday.

3. Write an email to your MP about an issue of

concern to you at the present time

Functional writing task

• Participants completed at following time points:

• Each email timed and 3 minute cut off

• All video-recorded

• Discourse analysis measures:

– Number of units

– Number of correct units

– Number of correct and informative content words

With

out Co:Writer

With

Co:Writer

Pre

effort phase

 

Post effort phase

 

Post therapy

 

6 month

follow-up

 

Control participants

10 healthy control participants

Healthy controls

PWA

Age

Mean

62.1

60.4

11.1

12.0

Education

Mean

12.1

11.3

2.2

2.1

Control data

Items

Correct items

Correct and

informative

items

Mean

174.5

173.1

101

94.9

56.7

• Each completed email writing task

• Maximum 3 minutes per email

• Data for combined tasks:

Co:Writer 7 (Don Johnston Technologies)

Word prediction

Co:Writer 7

Spell check

Co:Writer 7

Sensitive to regularisation errors

Co:Writer7

Grammar prediction

Co:Writer 7

Word/ phrase banks

Co:Writer 7

Text to speech

Co:Writer 7

Can use online (e.g. email, Facebook, Skype) – in theory!

Therapy protocol

x10 - 1 hour sessions over 5 weeks

Each therapy session consisted of

• Technology access training (0-15 minutes)

• Writing with Co:Writer (45-60 minutes)

Technology Access Training

Keyboard practice: copying short texts

Email training:

1. Turn on the computer

2. Open your email account

3. Create a new message, enter an e-mail and send a blank e-mail.

4. Send an e-mail with an attachment

5. Send an e-mail with a subject entered

6. Reply to an e-mail in your inbox

7. Enter an addressee into the cc box

8. Forward an e-mail

Finish when each task has been completed independently x3

Writing with Co:Writer

Session

Topic

Level

Introduction

& orientation

Writing

for domestic needs

Simple

tasks

Writing

for social needs

Simple

tasks

Writing

for business/ administrative needs

Simple

tasks

Writing

for domestic needs

Medium

complexity

Writing

for social needs

Medium

complexity

Writing

for business/ administrative needs

Medium

complexity

Writing

for domestic needs

High

complexity

Writing

for social needs

High

complexity

Writing

for business/administrative needs

High

complexity

Write a birthday card to a friend

Instructions to your neighbour, explaining

what they need to do in your house while

you’re on holiday.et).

Write to a charity/ organisation/

company explaining why you would

like to work/ volunteer for them and

why they should employ you

General principles

• At beginning of session entered words and phrases into word

banks together

• Encouraged to use word bank, prediction, text to speech and

spell check to complete tasks

• Participants completed tasks as independently as possible

• Could ask for therapist’s help (modelling, instructions)

whenever necessary

• Possible 3 tasks to complete per session, but worked through

at own pace

• Continued to next topic/ difficulty level in next session.

Age

Ed.

LR/R

Time since

CVA

Aphasia

Dysgraphia

years

Anomic

GBD

3 years

Non

-fluent

Phonological

7 years

Non

-fluent

Phonological/deep

6 years

Non

-fluent

Deep

5 years

Conduction

GBD

19 years

Non

-fluent

GBD

4 years

Recovered

Surface

4 years

Fluent with

phonological

errors

GBD

Age

Ed.

LR/R

Time since

CVA

Aphasia

Dysgraphia

years

Anomic

GBD

3 years

Non

-fluent

Phonological

7 years

Non

-fluent

Phonological/deep

6 years

Non

-fluent

Deep

5 years

Conduction

GBD

19 years

Non

-fluent

GBD

4 years

Recovered

Surface

4 years

Fluent with

phonological

errors

GBD

Age

Ed.

LR/R

Time since

CVA

Aphasia

Dysgraphia

years

Anomic

GBD

3 years

Non

-fluent

Phonological

7 years

Non

-fluent

Phonological/deep

6 years

Non

-fluent

Deep

5 years

Conduction

GBD

19 years

Non

-fluent

GBD

4 years

Recovered

Surface

4 years

Fluent with

phonological

errors

GBD

Age

Ed.

LR/R

Time since

CVA

Aphasia

Dysgraphia

years

Anomic

GBD

3 years

Non

-fluent

Phonological

7 years

Non

-fluent

Phonological/deep

6 years

Non

-fluent

Deep

5 years

Conduction

GBD

19 years

Non

-fluent

GBD

4 years

Recovered

Surface

4 years

Fluent with

phonological

errors

GBD

Age

Ed.

LR/R

Time since

CVA

Aphasia

Dysgraphia

years

Anomic

GBD

3 years

Non

-fluent

Phonological

7 years

Non

-fluent

Phonological/deep

6 years

Non

-fluent

Deep

5 years

Conduction

GBD

19 years

Non

-fluent

GBD

4 years

Recovered

Surface

4 years

Fluent with

phonological

errors

GBD

Age

Ed.

LR/R

Time since

CVA

Aphasia

Dysgraphia

years

Anomic

GBD

3 years

Non

-fluent

Phonological

7 years

Non

-fluent

Phonological/deep

6 years

Non

-fluent

Deep

5 years

Conduction

GBD

19 years

Non

-fluent

GBD

4 years

Recovered

Surface

4 years

Fluent with

phonological

errors

GBD

Baseline CAT writing scores

Participants:

CAT Scores (no.

letters correct)

Copying

27/27

25/27

26/27

27/27

Written

picture

naming

7/21

13/21

19/21

17/21

13/21

17/21

18/21

Writing

dictation

12/28

5/28

17/28

6/28

13/28

16/28

26/28

24/28

Written

picture

descripti

Results

• Across all 3 emails : Tasks 1, 2 and 3 collapsed

• Time limit 3 minutes for each email – 9 minutes total

Number of tasks completed independently in the

Internet Skills Assessment

LR GP DM KR AD JB SR EB Mean

Number of tasks completed

independently

Participant

Figure 1. Number of tasks completed independently in the Internet

Skills Assessment

Pre therapy Post therapy

p < .05

LR GP DM KR AD JB SR EB Mean

Number of tasks completed independently

Participant

Pre therapy Post therapy

p < .05

Speed of typing within Keyboard Skills

Assessment

LR GP DM KR AD JB SR EB Mean

Number of minutes for all tasks

Participant

Pre therapy Post therapy

Effort Phase

Mean

Pre effort

phase

Post effort

phase

Mean

Does email writing improve with effort alone?

Correct units per minute Correct and informative units per minute

Writing without Co:Writer

M…

Pre therapy

Post therapy

M…

Correct units per minute Correct and informative units per minute

Are there any general improvements to email writing without using

Co:Writer?

LR GP DM KR AD JB SR EB Mean

Correct units

Participant

Number of correct units pre therapy without Co:writer

compared to post therapy with Co:Writer

pre therapy without

Co:Writer

post therapy with Co:Writer

* p < .05

LR GP DM KR AD JB SR EB Mean

Correct and informative units

Participant

Number of correct and informative units pre therapy

without Co:Writer compared to post therapy with Co:Writer

pre therapy without

Co:Writer

post therapy with

Co:Writer

*p < .05

LR GP DM KR AD JB SR EB Mean

Rating

Participant

CAT Disability Questionnaire:

Ratings on writing subtest

Pre therapy Post therapy

*p < .05

Proportion (%) of each word class in pre and post therapy emails

Participant

Adj

Adv

Exclam

Num

Function

Pre

Post

Pre

Post

Pre

Post

Summary

Engaging in this treatment programme meant that:

o participants became more independent though not faster in

internet skills

o Effort and non-use of the compensatory technology showed

no gains for email writing

o Group level gains in correct words and correct/informative

words with compensatory technology, some greater variety of

word classes used

o This effects were carried by half of the participants

o Perceived impact on functional writing skill

Factors affecting individual variation

Cog/ ling

Prior use

Spelling type/ severity

Motivation

Ease of use of technology

Future directions

 Investigation of relationships between performance on

cognitive/ language/ spelling assessments and outcomes

 Prescriptive intervention – what should work best for you

is…..

 Interviews and analysis of interview data

 Aphasia-friendly app

 Literacy focus - combined reading & writing package

 Quality of written language production

• Thiel, L., Sage, K. & Conroy, P. (2013). Comparing uni-modal and

multi-modal therapies for improving writing in acquired dysgraphia.

Poster presented at the British Aphasiology Society Biennial

International Conference 2013, University of Manchester.

• Thiel, L., Sage, K. & Conroy, P. (2013). Combining word learning

therapies and computer technologies to improve writing in people

with aphasia. Paper presented at the Speakeasy Annual Conference,

Bury, Greater Manchester.

• Thiel, L., Sage, K. & Conroy, P. (2014). The role of learning in

promoting functional writing. Poster presented at the 16th

International Aphasia Rehabilitation Conference 2014, The Hague,

The Netherlands.

• Thiel, L., Sage, K. & Conroy, P. (2014). Training people with aphasia

to use word prediction software for email writing. Paper presented

at the British Aphasiology Society Therapy Symposium, Birmingham

City University.

• Thiel, L., Sage, K. & Conroy, P. (2015). Retraining writing for

functional purposes: a review of the writing therapy literature.

Aphasiology, 29(4), 423-441.

Discussion

DM, EB and GP

• Did not improve significantly;

• What do they have in common?

– All mastered Co:Writer very quickly, enjoyed using it and wrote well

within therapy. DM and EB continued using after therapy.

– Were writing emails before and during study

– Good cognitive skills

• Possible reasons for no improvements

– Chose to use predictive writing more than word banks = slow to use

and don’t see such large improvements in correct words per minute

– Assessment not sensitive to change

Discussion

AD, JB and LR

• All improved significantly when using Co:Writer post therapy

• What do they have in common?

– Severely impaired spelling; graphemic buffer disorder

– Severely impaired email writing and dependent on support

– Severely impaired cognitive skills (executive functioning; visual memory)

• In what way did Co:Writer help them?

– All relied on word/ phrase bank function – one click produces multiple

words.

– However, all have extreme difficulties in using Co:Writer (found it

frustrating) within therapy and none used Co:Writer independently after

therapy