Development of Assistive Technology
When Learning Disability is no Barrier
Onintra Poobrasert and Waragorn Gestubtim
National Electronics and Computer Technology Center, National Science and Technology Development Agency,
Pathum Thani 12120, Thailand
Keywords: Assistive Technology, Dysgraphia, Learning Disabilities, Soundex, Spell Checker, Text-to-Speech.
Abstract: Assistive technology does not cure a specific learning disability. These tools are allowing a student with
learning disabilities to demonstrate their intelligence and knowledge. Assistive technology for the person
with learning disabilities is a made-to-fit implementation. Trial and error may be required to find a set of
appropriate tools and techniques for a specific individual. Therefore this paper is aimed to attest the use of
assistive technology: Thai Spell Checker program which enhanced students with learning disabilities in their
writing. This paper will also depict the details of the design and development of assistive technology tool,
including information on the experiment with real users. The results from the study show that the assistive
technology tool has a correct rate of 89% for detecting misspelling words. Additionally results indicate that
the students have writing improvement in Thai language as 75%, 15%, 30%, 70%, and 35%.
1 INTRODUCTION
Assistive technology will be a tool for enhancing
capacity or adjusting ability of a special child and
will allow such child to learn better and more
efficiently. In addition, Assistive technology is a
technology that instructor can apply for facilitating
the learning activity and increasing learning
opportunity (Alper and Raharinirina, 2006).
Learning methods of students with learning
problems are different from other students. The
techniques used are also different as well as the
technologies. Although some students with learning
disabilities will be able to study with normal
students, they require special processes or methods
of teaching to draw out their expertise and unique
talents in order to replace or remove their
weaknesses (Gersten et al., 2001). In Thailand most
students with learning disabilities still have to learn
the same way as normal students because the use of
assistive technology in the country is still limited
(Poobrasert et al., 2011). Though in the market,
there are various types of assistive technology for
students with learning disabilities, those tools are not
suitable and not benefit Thai students. Most of this
software is designed and developed for English
speakers. The prices of those assistive technology
tools are also expensive. Since Thai language is very
unique and different from other languages, we then
needed to integrate Thai text to speech (VAJA) and
BEST corpus into the program. BEST corpus
consists of Thai vocabularies from article,
encyclopedia, news and novel. We also added Thai
vocabularies from Thai textbooks. Hence, we
have to design and develop our own assistive
technology for students with learning disabilities in
Thailand. As mentioned earlier, assistive technology
will help increase the ability or adjusting proficiency
of students with learning disabilities to learn
effectively.
According to Shaywitz (2005) children with
cognitive or learning difficulties need more general
solutions which include providing a consistent
design and using simplified language. They also
gain benefit and comprehend the material better
from viewing the text and hearing it read aloud.
Therefore it is very important for the researcher to
have a better understanding of disability issues.
Moreover, the key to accessible user interface and
the process of usability engineering are to ensure
accessible user experience and their limitations. In
this study, we also performed usability engineering
to ensure that the Thai Spell Checker program is
ready to use with the students with learning
disabilities in Thailand.
159
Poobrasert O. and Gestubtim W..
Development of Assistive Technology - When Learning Disability is no Barrier.
DOI: 10.5220/0004719201590165
In Proceedings of the International Conference on Health Informatics (HEALTHINF-2014), pages 159-165
ISBN: 978-989-758-010-9
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
2 RELATED RESEARCH
2.1 The Thai Word Search Function
The Thai Word Search program is an assistive
technology tool for students with learning
disabilities in writing by providing help in searching
for vocabulary either in the mode of Homophony or
Soundex (a word that is pronounced the same but
has different grapheme) or Word Approximation ;
words with a similar depiction; probably be
misspelled or wrong tone (Wren, 2004).
With respect to Thai Word Search program, when
the student wants to write any vocabulary that he/she
cannot spell it correctly, the student will just type
part of word according to pronunciation or as
guessed, the program will then check and
demonstrate the words for selection that most likely
matches the one desired by the student including
their pronunciation.
2.2 Thai Text to Speech
In this paper, we applied the use of Thai text to
speech (VAJA: Research and development of
Speech and Audio Technology Lab [SPT],
NECTEC) (SPT Lab, 2013) into our program.
Text to speech is a technology that can create
any voice as desired and most of its use has to be
associated with language processing technology.
Text to speech can be integrated into Thai text in
order to read and convert voice into it in Thai for
every word as it has the function for word
pronunciation, even for a word not appeared in a
dictionary. In addition, the user can add a specific
word such as an individual’s name and freely
customize pronunciation in order that software can
convert text into voice as desired by the user.
Our program, Thai Spell Checker applies Thai
text to speech as one of its components aimed to
read text and convert it to voice for the user.
Pronunciation of vocabulary by Thai Spell Checker
can help students with learning disabilities to hear
and select the desired vocabulary correctly.
2.3 Thai Query Correction
Another previous technology is called Thai Query
Correction. Thai Q Corr is a service rendered in the
pattern of web server with an ability to verify
vocabulary wrongly written by the user, by
presenting in, for instance, homophony or related
written word due to wrong word writing or spelling
(Haruechaiyasak et al., 2008).
The Thai Spell
Checker program also applied the use of “Linguistic
Rule for LD” into the program.
3 DESIGN AND DEVELOPMENT
OF ASSISTIVE TECHNOLOGY
3.1 Program Design
The main system architectural design of the Thai
Spell Checker program is shown in Figure 1.
Figure 1: Main System architectural design of Thai Spell
Checker Program.
Program design consists of searching function.
The function can search for Homophony or Soundex
and Word Approximation (see Figure 2).
3.1.1 Homophony or Soundex Search
Homophony or Soundex search will take searching
word from the user and then apply W2P (Word-to-
Phoneme; the phoneme is the basic building block
for spoken words) tool to convert searching word
into voice through the converting process which will
discard ร, ล (two Thai letters having close
pronunciation) before searching into a hash database
of voice code, to gain the result in writing pattern
and return it to the user. Homophony or Soundex
search will pay attention to the only pronunciation of
such word. When we have got a word in the writing
pattern (Grapheme) we will then find its
pronunciation and it is called Phoneme. Normal
grapheme will be definite while Phoneme may be
Phonetic or code or other kind of voice
(Angkawattanawit et al., 2008).
HEALTHINF2014-InternationalConferenceonHealthInformatics
160
Figure 2: The system architectural design of Thai Word
Search Function.
Figure 3: G2P of โรงเรียน and อากาศ.
Procedures for Homophony or Soundex Search: It is
able to convert Grapheme into Phoneme more
easily. This will start from dividing words into
syllables via the process of Lexeme Tokenizer (Su
and Song, 2011) applying a mark “|” to define
syllable. The intonation mark of such individual
syllable will be then deleted by a Tone Remover
process to avoid being confused by the system.
Cancelled Letter Remover will be applied to delete
the mark from syllable before the last process of
converting Grapheme into Phoneme (G2P)
(Kasuriya et al., 2003) (see Figure 3).
Data storage in Homophony or Soundex search
system is done by the hash database (Main and
Savitch, 2010). The advantage of hash database is
the focus on swift searching and adding/deleting
member until the static time O (1) but such data has
to have no priority and duplicity (Viola, 2005). This
first step is to take word aimed to search for into the
searching system G2P by following the process
mentioned above and voice code from these words
will be converted where
ร, ล will be discarded by
the system (see Figure 4).
Figure 4: Sample of words where ร, ล were discarded by
the system.
When these phonemes are identified, searching
process will continue under the following steps:
- All correct words search,
- 1 wrong word searches which will occur when
the wrong word entered by the user has more
than 2 syllables, and
- All words in the vocabulary search pattern
These 3 processes will be followed respectively and
if the result is gained at any step, it will be presented
immediately.
3.1.2 Approximation Search
Search for approximation means searching for words
having close Grapheme to designated word and may
contain slightly different letter e.g. “
กาลเทศะ” –
“
กาลเทศะ” or “วิจารณญาน” –
“
วิจารณญาณ” etc. These samples are set of
homophony which can be searched by homophony
or soundex search system. From the other
perspective, these words contain only few different
letters e.g. “
วิจารณญาน” – “วิจารณญาณ”
differ from each other by using “
น” or “ณ”. This
approximation search will ignore pronunciation but
close grapheme (Poobrasert, 2011). By the use of
approximation search, the key attention is mistyping
e.g. “
ผจยภัย” – “ผจญภัย”, “กรุงศรีอยุธบา” –
“
กรุงศรีอยุธยา” or “กลลวง – กลลงว” etc. It
can be noticed that some word cannot be even
spelled out or can be spelled without meaning. In
case of “
ผจยภัย” – “ผจญภัย”, it may be the
result of mistyping as the letter “
ญ” and “ย” are on
the same key but the shift key has not been pressed
properly to change from “
ย” to “ญ”.
Procedures for Approximation Search: Data
storage in approximation search system is done
by hash database. The advantage of which is the
focus on swift searching and adding/deleting
DevelopmentofAssistiveTechnology-WhenLearningDisabilityisnoBarrier
161
member until the static time O (1) but such data
has to have no priority and duplicity.
First step is to take designated word to be processed
under the steps as follows:
- Search for all correct words, search by the Rule of
LD (Pothibal, 2010) 1
st
round by the method
shown below:
o Deletion: refer to the step that word typed by
the user will be deleted letter by letter, the
result is n.
o Transposition: refer to the step that word typed
by the user will be switched letter by letter, the
result is n-1.
o Substitution: refer to the step that word typed
by the user will changed letter by letter under
Rule of LD e.g. letter “
ก
” can be changed to
“
ก
” “
ภ
” “
ถ
” “
ฤ
” “
ด
” “
บ
” “
ม
” or “
ย
”.
o Insertion: refer to the step that the word typed
by the user will be added letter by letter, the
result is (n+1) x 66 results.
- Search again by Rule of LD, 2
nd
round.
The above processes will be followed step by step
and if the result
is gained at any step, it will be
presented immediately. However, this method has
some limitation, that is, it cannot correct the word
containing fault for more than 2 letters and thus API
Approxy, Edit distance has been introduced to
increase the efficiency for search.
Figure 5 shows the procedures for approximation
search.
Figure 5: Procedures for approximation search.
Interface Design of Thai Spell Checker Program:
Figure 6 demonstrates the interface design of the
Thai Spell Checker Program. The user uses the
ribbon to issue the commands within the
program. At the top of the ribbons are several
tabs such as “File”, “Format”, “Insert” and
“Setting”. Spell Checker Button is located in the
File tap.
Figure 6: Interface design of Thai Spell Checker Program.
When click the Spell Checker button, the new
dialog box will appear. This box details the spell
checker working area (Figure 7), including
misspelled words which are indicated by pink
highlight and suggested word lists.
Figure 7: The Thai spell checker working area.
4 METHODOLOGY
The usability testing was performed before
conducting the experiment with real users. Usability
testing measures the usability of the system
(Cockton, 2013). In this study we use heuristic
evaluation (Hvannberg and Lárusdóttir, 2007) which
is the most popular of the usability inspection
methods (Nielsen, 1994). Nielsen and Landauer
suggested that the best results come from testing no
more than five evaluators and running as many small
tests as we can afford (Rosson and Carroll, 2010).
HEALTHINF2014-InternationalConferenceonHealthInformatics
162
We are trying to fix all bugs and errors found
during the usability testing. At present the system
evaluation shows the correction’s scores of 89%. We
therefore expect that our assistive technology tool
will assist students with learning disabilities who
struggle with writing become better writers. Thus
the program was suitable to use in an actual
experiment.
In this study single subject research design (a
single-case experimental design) (Gay and Airasian,
2003) was also applied into our experiment. We use
this method when the sample size is one or when a
number of individuals are considered as one group.
These designs are typically used to study the
behavioral change an individual exhibits as a result
of some treatment. In single-subject design, each
participant serves as his or her own control. The
Thai Spell Checker program was experimentally
applied to five students with learning disabilities
who struggle with writing in grade 5
th
at the school
in Bangkok.
Method for Selecting Students:
Students with learning difficulties who struggle
with writing in grade 5
th
were selected to conduct
Pretest.
Students’ writing ability was tested on the
specific vocabulary where each of which was
allocated about 12 minutes to finish the test (25
words).
The researcher collected writing results from the
student and then selected five out of them (under
the assistance from the Thai instructor).
5 EXPERIMENT
5.1 The Experiment – Practicing
Method
All of the five students are assigned to write 20
vocabularies on their workbook without using the
assistive technology program. When finished they
are allowed to type the same set of vocabulary on
the computer and use the program “Thai Spell
Checker” to assist them. Figure 8 shows an example
of one LD student handwriting and Figure 9
demonstrates the Thai Spell Checker program.
Figure 8 shows an example of one LD student
handwriting and Figure 9 demonstrates the Thai
Spell Checker program. Figure 8 shows an example
of one LD student handwriting and Figure 9
demonstrates the Thai Spell Checker program.
Figure 8: Example of one LD student handwriting.
Figure 9: The Thai Spell Checker Program.
6 RESULTS
The results in Table 1 showed that students with
learning disabilities in this study improved their
ability of writing when use assistive technology.
The student#1 received the lowest score on his
pretest. With his endurance and trying to use
assistive technology to complete the tasks, he
received full score on his post test. Therefore his
improvement of writing was 75% which was the
highest score.
The student#2 received 13 scores on his pretest
and 16 scores on his post test. His writing
improvement was 15%.
The student#3 received 14 scores on her pretest
and 20 scores on her posttest. Consequently, the
improvement in learning of the students#3 was 30%.
The student#4 received seven scores on her
pretest and 20 scores on her post test. It is indicated
that her improvement in learning was 70%.
The student#5 received 13 scores on the pretest
and 20 scores on the post test. Hence, the assistive
DevelopmentofAssistiveTechnology-WhenLearningDisabilityisnoBarrier
163
technology helped him to improve his writing for
35%.
Table 1: Comparison of test scores.
Students WithoutAT WithAT
1 5 20
2 13 16
3 14 20
4 7 20
5 13 20
Figure 10: The Writing Improvement of all five students.
7 CONCLUDING REMARKS
This paper presents an assistive technology tool to
help students with learning disabilities in Thailand to
learn to write. The method addresses the
combination number of different word search
techniques to spell check a document and provide
interactive feedback to the users. Five students with
learning disabilities who participated in the study
were in agreement with the advantages of the
assistive technology; Thai Spell Checker which
could help Thai spelling and search Thai words well
and fast. The students mentioned that the program
assisted them to be able to choose vocabulary and
print the work correctly. The program also allowed
them to learn the meaning of new vocabularies.
They agreed that Thai Spell Checker was user
friendly, and not complicated.
In addition, the program also helps the students
to know how each word is spelled as it provides
pronunciation for each word. Although the
pronunciation by the program for some word was
deviated from the computer-generated simulation of
human speech (VAJA-TTS), most students in this
study admitted that the program could help them to
be able to pronounce vocabulary better than without
using the program. Not only the pronunciation, but
we also expect that assistive technology tool such as
Thai Spell Checker will help enhancing the writing
skill of students with learning disabilities in
Thailand.
Finally, in this study, we have not conducted the
experiment using other Spell Checker programs with
these students for the purpose of the study is to assist
the students who struggle with writing become better
writers in Thai language. The study is the first of its
kind to be carried out in Thailand, however the
approach and method adopted for this problem can
also be transferred to other languages.
ACKNOWLEDGEMENTS
We would like to convey our thanks and
acknowledge the assistance of the Speech and Audio
Technology Lab (SPT), National Electronics and
Computer Technology Center (NECTEC), Miss
Wantanee Phantachart, Asst. Prof. Dr. Puttachart
Pothibal, Mr. Alongkorn Wongteeratana, the
director, the teachers and the students at the School
in Bangkok. Our thanks also extend to the Cluster
and Program Management Office (CPMO), National
Science and Technology Development Agency
(NSTDA) for funding the project.
REFERENCES
Alper, S. & Raharinirina, S., (2006). Assistive technology
for individuals with disabilities: A review and
synthesis of the literature. Journal of Special
Education Technology, 21(2), 47-64.
Angkawattanawit, N., et al., (2008). Thai Q-Cor:
Integrating Word Approximation and Soundex for
Thai Query Correction, Proceedings of ECTI-CON.
Cockton, Gilbert, (2013). Usability Evaluation. In:
Soegaard, Mads & Dam, Rikke Friis (Eds.). "The
Encyclopedia of Human-Computer Interaction, 2
nd
Ed.". Aarhus, Denmark: The Interaction Design
Foundation. Available online at http://www.interaction
design.org/encyclopedia/usability_evaluation.html.
Gay, L. R., & Airasian, P., (2003). Educational Research:
Competencies for Analysis and Applications. Merrill
Prentice Hall: Columbus, OH.
Gersten, R., Fuchs, L., Williams, J., & Baker. S., (2001).
Teaching Reading Comprehension Strategies to
Students with Learning Disabilities: A Review of
Research, doi: 10.3102/00346543071002279 Review
of Educational Research Summer 2001 vol. 71 no. 2
279-320.
Haruechaiyasak, C., et al., (2008). A Comparative Study
on Thai Word Segmentation Approaches, Proceeding
of ECTI-CON.
Human Language Technology Lab., (2009). HLT-VAJA
Retrieved [September 2013] from the World Wide
HEALTHINF2014-InternationalConferenceonHealthInformatics
164
Web: spt.nectec.or.th/spt/.
Hvannberg, E., Law, E., & Lárusdóttir, M. (2007).
Heuristic Evaluation: Comparing Ways of Finding and
Reporting Usability Problems, Interacting with
Computers, 19 (2), 225-240.
Kasuriya, S., Sornlertlamvanich, V., Cotsomrong, P.,
Kanokphara, S., & Thatphithakkul, N., (2003). Thai
Speech Corpus for Thai Speech Recognition, The
Oriental COCOSDA 2003, 54-61.
Main, M., & Savitch, W., (2010). Data Structures and
Other Objects Using C++. 4
th
Ed., Prentice Hall.
ISBN 978-0132129480.
Nielsen, J. (1994). Heuristic evaluation. In Nielsen, J., &
Mack, R.L. (Eds.), Usability Inspection Methods. John
Wiley & Sons, New York, NY.
Poobrasert, O., et al., (2011). Usability in Designing
Assistive Technology for Children with Learning
Disabilities. Proceeding of the 5
th
International
Convention on Rehabilitation Engineering & Assistive
Technology, iCREATe 2011, Thailand.
Poobrasert, O., et al., (2011). Technology-enhanced
Learning for Students with Learning Disabilities.
Retrieved [October 2011] From the World Wide Web:
http://doi.ieeecomputer-society.org/10.1109/ ICALT.2
011.154.
Pothibal, P. (2010). Report on Variants of the Word is
used as a Rule of LD. (Unpublished).
Rosson, M. & Carroll, J. (2010). Scenario-Based Usability
Engineering (Paperback). Morgan & Claypool
Publishers. San Rafael, CA.
Shaywitz, Sally. (2005). Overcoming Dyslexia, A New and
Complete Science-Based Program for Reading
Problems at Any Level. 1
st
Ed., Vintage. ISBN-13:
978-0679781592.
Su, Yunlin., & Song Y. Yan. (2011). Lexical Analysis
Principles of Compilers. Springer Berlin Heidelberg.
DOI 10.1007/978-3-642-20835-5_4.
Viola, A. (2005). Exact Distribution of Individual
Displacements in Linear Probing Hashing.
Transactions on Algorithms (TALG) (ACM) 1 (2,):
214–242.
Wren, S. (2004). Phoneme Awareness. Retrieved [August
2013] from the World Wide Web:
http://www.balancedreading.com/phonemeawareness.
html.
DevelopmentofAssistiveTechnology-WhenLearningDisabilityisnoBarrier
165
