Content uploaded by Martin Ebner
Author content
All content in this area was uploaded by Martin Ebner on Nov 27, 2015
Content may be subject to copyright.
eLearning
Papers
45
4
In-depth
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
Authors
Konstanze, Edtstadler
konstanze.edtstadler@kphvie.
ac.at
University College of Teacher
Education Vienna/Krems,
Researcher
Vienna / Krems, Austria
Markus, Ebner
markus.ebner@tugraz.at
Institute of Information
Systems and Computer
Media, Graz University of
Technology, Researcher
Graz, Austria
Martin, Ebner
martin.ebner@tugraz.at
Department Educational
Technology, Graz University
of Technology, Head of
Department
Graz, Austria
Improved German Spelling Acquisition through
Learning Analytics
Many pupils struggle with the acquision of the German orthography. In order to
meet this struggle a web based plaorm for German speaking countries is currently
developed. This plaorm aims to movate pupils aged 8 to 12 to improve their wring
and spelling competences. In this plaorm pupils can write texts in the form of blog
entries concerning everyday events or special topics. Since the core of this plaorm
consists of an intelligent diconary focussing on dierent categories of misspellings,
students can improve their own spelling skills by trying to correct their mistakes
according to the feedback of the system. Teachers are informed about specic
orthographic problems of a parcular student by geng a qualitave analysis of the
misspellings from this intelligent diconary. The arcle focuses on the development
of the intelligent diconary, details concerning the requirements, the categorizaon
and the used wordlist. Further, necessary informaon on German orthography, spelling
competence in general and the plaorm itself is given. By implemenng methods
of learning analycs it is expected to gain deeper insight into the process of spelling
acquision and thus serves as a basis to develop beer materials in the long run.
1. Introduction
This arcle is concerned with a learning analycs approach in the eld of German orthography.
Due to the increasing internet usage in the eld of educaon, the amount of data that is
produced is rising daily. This data is shared between various kinds of instuons around
the globe (Piety, 2013). Furthermore, the heavy use of the Internet generates enormous
data about learners’ behavior and leaves traces of every interacon (Duval, 2010). Thus,
interacon between students and a learning plaorm can be captured and used for later
analysis in order to gain an insight into a learners’ learning process (Khalil & Ebner, 2015).
This can then be used for early detecon of learning issues and enables teachers to acvely
intervene accordingly in order to solve such issues eecvely (Siemens et al., 2011; Greller
& Drachsler, 2012).
German orthography is known to be quite dicult to master. People from dierent social
classes, of dierent ages and with varying degrees of educaon, struggle with spelling words
correctly. However, the importance of correct spelling for social acceptance is quite high. It
aects primary-school pupils’ as well as a university students’ everyday life inside and outside
German orthography,
learning analytics, qualitative
analysis of misspellings,
categorization, Technology
Enhanced Learning
Tags
5
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
schooling. Sll, instrucons in German orthography are oen
unsystemac and not parcularly aracve for children.
The development of the IDERBLOG-Plaorm1 aims to solve
such problems by combining technology enhanced learning
and learning analycs with the acquision of German
orthography (Ebner et al., 2015a). The plaorm should serve
as an aracve and movang innovaon for children to
acquire German orthography appropriately and more easily.
It has also advantages for teachers and researchers, as the
applicaon of learning analycs supports them in their decision
making process by providing them with an overview of possible
educaonal intervenons (Ebner et al., 2015b).
Outline
The next secon gives a short overview of the German
orthography as well as orthographic competence and
learning analycs. The following secon is concerned with the
development of the informaon system of the plaorm, its
interface design process and the planned workow. The two
succeeding secons focus on to the intelligent diconary and
the feedback system. The arcle aims to give an overview of the
categories, the requirements and the process of categorizaon
of the intelligent diconary.
2. Related work
German Orthography
German orthography uses an alphabec wring system.
Alphabec wring systems are characterised by mirroring the
phonemic structure of the spoken language to the wrien
language, which leads to the assumpon that words are spelled
as they sound (cf. Katz & Frost, 1992, p. 149).
This phonological principle is applied to a varying degree of
consistency in dierent languages. It leads to a connuum
of orthographies ranging from transparent to opaque ones
with a huge impact on spelling instrucon and acquision. In
transparent orthographies like Serbian, Turkish or Italian each
phoneme (notated consecuvely with / /) is represented by one
leer – or more precisely – grapheme (notated consecuvely
with < >). Therefore, the assumpon to spell a word as it is heard
is quite true in these orthographies. In opaque orthographies
like English or French “there is a pronounced discrepancy
1 IDERBLOG-Plaorm, available online: hp://iderblog.eu/ (German language
only, last visited October 9, 2015)
between pronunciaon and orthography” (Klees, 1989, p.
137). Consequently, learners are confronted with unreliable
correspondences since – in the case of English for example –
“the alphabet contains just 26 leers yet these correspond to
44 phonemes associated with 102 funconal spelling units.”
(Snowling, 1989, p. 1). The German orthography can be found
in the middle between transparent and opaque orthographies.
Following Nerius (2007) it consists of two basic principles, the
phonological and the semanc principle. Part of the phonological
principle are the phoneme-grapheme-correspondences (PGC),
which are mostly not in a 1:1 relaonship, e.g. /a:/ can be <a>
in <Wal> whale, <aa> in <Saal> hall or <ah> in <kahl> bald (cf.
Meinhold & Stock, 2007, p. 122). Part of the semanc principle is,
according to Nerius (2007, p. 89 ), the morphological principle
- among the lexical, the syntacc and the textual one. This
principle is responsible for spelling one morpheme in the same
way in all words in which it occurs. This oen leads to a conict
with the phoneme-grapheme-correspondences: e.g. spelling the
word dog, in German pronounced as /hunt/, following the PGC
would lead to the misspelling (usually indicated with an asterisk)
<*Hunt>. It has to be spelled <Hund> because of the plural form
/hundə/ dogs. The spelling <Hund> with a <d> is kept the same
in all words, like <Hündin> female dog or <Hundeschlien> dog
sled. The spelling of the orthographically challenging ‘Umlaut’
(= vowel mutaon spelled as ä/äu) in morphologically complex
words is also due to the morphological principle (e.g. <Hände>
hands, not <*Hende>).
These principles and their value for German orthography
highly aect the didacc approach of teaching – especially in
higher classes and addional trainings. In general, the spelling
instrucon at the beginning of literacy acquision is clearly
phoneme based (cf. Landerl & Thaler, 2006). This is the reason
why children are able to write (new) words relying on their
knowledge of pronunciaon soon aer geng to know the
PGC. Words wrien in this way can also be read, but they are
oen not orthographically correct (e.g. <*falipt> for <verliebt>
in love). Especially for children who are not speaking the
standard German variety, the inuence of the spoken language
is evident in their spelling. Due to other sub-principles of the
semanc basic principle further orthographic challenges are –
for example:
• Nouns must be spelled with capital leers – a feature that
can only be found in the German orthography (cf. Valn,
1989, p. 119). It leads to many mistakes – even in texts of
well-educated adults.
6
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
• Homophone words are somemes, but not always, spelled
dierently (e.g. /li:t/ as <Lied> song or <Lid> eyelid, but /
notə/ as <Note> for mark and note) (cf. Nerius, 2007, 167).
• Compounds are usually spelled compound (e.g. <Teetasse>
tea cup). Depending on the meaning of a combinaon of
words, it must be spelled separately or compound (e.g.
<Schweinebraten> roast pork or <Schweine braten> – to
fry pork) (cf. Fuhrhop, 2011, p. 107).
Spelling Competence
Especially in the German speaking world correct spelling is
considered very presgious, but students consider spelling
instrucons oen as boring and formal (cf. Küel, 2007, p.
418f). Addionally, teachers oen do not pay aenon to the
systemac principles that stand behind the spelling of certain
words. This oen leads to the assumpon, that it is not possible
to teach German orthography systemacally (cf. Fröhler, 2002).
In contrast to other areas of language learning, there is hardly
space to argue about the correct or incorrect spelling of a
word. This orthographical sness can probably serve as an
explanaon for its importance.
It is important to understand that the spelling competence of
a person does not only include the knowledge of the correct
spelling of a given word and knowing the rules of orthography.
It also includes being sensive to misspelled words, knowing
how to correct them, using spelling aids and applying strategies
(cf. Sommer Stumpenhorst, 2012; Naumann, 2008). Concerning
instrucon, it is not enough to simply oer dierent online or
oine exercises. “Children or student’s need purposeful reading
and wring in a broad range of situaons, in an environment
that values risk-taking. They will develop spelling competence
as they implement their knowledge of the spelling system,
receive feedback and rene their hypotheses.” (Government
of South Australia, 2011, p. 6). Furthermore, children should
be encouraged to think about and reect language in order
to become aware of the structure of words (cf. e.g. Tsesmeli
& Seymour, 2006). Due to the dierent principles of German
Orthography, metalinguisc awareness must be established
beyond phonological awareness (cf. e.g. Naumann, 2008). For
example, children must be encouraged to see the morphological
link between singular and plural form (e.g. <Hälse> because of
<Hals> necks, <Rind> because of <Rinder> cows).
Learning Analytics
The eld of Learning Analycs tries to consider the learning
process as a whole in its full complexity. According to Baker et al.
(2012) and Neuhold (2013) it is important to keep feedback and
its visual representaon as simple as possible to avoid confusion
and unreasonable interpretaon on the side of the stakeholders.
Campbell et al. (2007) provide a model for the analysis process
in ve steps: capture, report, predict, act and rene. Clow
(2012) used these ve steps as a basis for his learning analycs
cycle. This iterave process consists of four main components:
learners, data, metrics/analycs and intervenon (Clow, 2012).
To get an overview about the whole process Khalil & Ebner
(2015) added stakeholders to the cycle. Nevertheless, the main
idea of Learning Analycs is to provide and process a learners’
data in an appropriate way in order to facilitate teachers to
react and (if necessary) to intervene. For instance, Taraghi et al.
(2015a) introduced an analycal approach to model a learner’s
prole according to their answering behavior. Moreover, the
analysis of dierent error types can lead to ndings that help to
enhance the learning process as a whole (Taraghi et al., 2015b).
3. Information system
The plaorm (informaon system) for the project is
currently under development and yet not available for public
presentaon. Nevertheless, this secon will provide basic
design ideas to ensure good age-appropriate interface design
and usability (Ebner et al., 2015a). In the second secon the
planned workow of the analysis will be outlined.
Writing by using the Computer
Since developing wring skills and acquiring orthographic
competence is important and wring with computers is
aracve for children, the IDERBLOG-Plaorm combines these
components. The aim is not to replace handwring by typing on
keyboards, but to take advantage of the digital age. “For some
people with major handwring problems, personal computers
are a boon.” (Høien & Lundberg, 2000, p. 68)
A further advantage of wring on a computer is, to train the
ability to correct texts. Since correcons are made within a digital
text, correcons do not leave traces in contrast to a handwrien
text. Consequently, a text can be edited several mes unl it
becomes publishable. Furthermore, the IDERBLOG-Plaorm
is “providing relevant reasons and audiences for wring”
7
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
(Government of South Australia, 2011, p. 8) as children can
publish their texts. Therefore, it is expected that the movaon
to formulate a text and to revise it several mes is possibly
higher in contrast to typical essay wring in a classroom where
the addressee is almost only the teacher.
Concerning the training of orthographic skills, the IDERBLOG-
Plaorm oers an intelligent diconary, which does not only
count the number of mistakes in a text, but also categorizes
the mistakes in dierent orthographic areas. In contrast to the
work of Thelen (2010), that analyses misspellings in German
orthography, we do not only focus on beginning spellers but
also on more advanced learners. One of the most important
features of the intelligent diconary is that it oers feedback
and hints for the correcon of a mistake. Addionally, the
plaorm oers a number of exercises that are connected and
categorised according to spelling mistakes and therefore meet
the need of pracce in a specic area of spelling.
Interface Design
The plaorm is generally designed for children the primary
school (age 8 to 12) with the focus on a graphically appealing
and age-appropriate web interface (Liebal et al., 2011). For
this purpose, a graphic designer created dras that have been
examined and rated by students from dierent schools. The
designs that were favoured by the majority were then, in a
second step, developed further and aerwards integrated into
the plaorm.
Another important part of the plaorm is usability. We had to
ensure that the students can reach the most important parts of
the plaorm in less than ve clicks. This convenient accessibility
in combinaon with aracve gures should ensure high
movaon in fullling the task of wring texts. In ongoing
usability tests (Holzinger et al., 2005) we connue to improve
the concept step by step.
Workow of the Platform
The students, as shown in Figure 1, can write their texts on the
provided plaorm. First the text will be analysed orthographically
by the intelligent diconary (which will be described in the
next secon). Proper feedback will be provided to the student,
based on error type and category. The student has the choice
to either try to correct the wrong words or to hand-in the text
directly to the teacher. This intermediate step encourages the
experse of independent correcon (Bartnitzky et al., 2010).
Aer the submission, the teacher should inspect the text for
further correcons and/or improvements. Notes can be made
and delivered with the nal correcon to the student. Aer this
step the text can be published in the class blog of the school (if
appropriate).
Figure 1: Workow of the plaorm
The methods of learning analycs will be used for further
analysis of the texts (Siemens, 2012). The results will be
provided to students, teachers and parents in an appropriate
way. Further, an overview of the frequency of mistakes and
possible systemacally made errors is oered. In the long run
changes in a students’ performance will be measured (Schön et
al., 2012).
Training database
The plaorm will include an addional training database, as
shown in Figure 1, with selected online exercises and oine
work sheets. This database will aid teachers and students
to nd appropriate exercises to improve the performance in
problemac areas (as a consequence of the learning analycs
analyses). The exercises and work sheets are congruently
ordered in categories and sub-categories for easier selecon.
4. The intelligent dictionary
Categorization of mistakes
A word can be either orthographically right – in case the spelling
of a given word exists in a list of correctly spelled word, called
8
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
diconary, or wrong – in case it does not exist. In science,
mistakes are analyzed in dierent categories depending on the
purpose of the study: e.g. for the English language Broc et al.
(2013) categorise spelling errors of people with specic language
impairments in phonologically acceptable vs. unacceptable. Flor
& Futagi (2012) focus on non-word misspellings in the context of
spell checker. In school oen a quantave approach is applied,
which means counng the number of misspelled words. In
addion to the correct-wrong dichotomy there are some other
ways in the categorisaon of incorrectly spelled words that lead
to a greater insight into the orthographic competence.
One way is to count the number of correctly wrien graphemes
of a given word: This helps to analyse the progress of extremely
weak or very young spellers (cf. May, 2010). For example, the
spelling for /V/e/r/k/äu/f/e/r seller in *F/e/r/k/eu/f/a contains
only 4 out of 8 correct graphemes in contrast to */V/e/r/k/
eu/f/e/r with 7 out of 8 correct graphemes. Although both
wring aempts are wrong, the second one is much beer.
This grapheme-based approach is a rather detailed and me
consuming way of correcng. Therefore, it is generally only
applied in a standardized spelling test called “Hamburger
Schreibprobe” (May, 2010) which provides pre-dened
templates for the quite small amount of words used in the test.
Another way of categorizing incorrectly spelled words is to dene
the type of mistake(s) and to collect the various frequencies
for the given categories in order to idenfy the orthographic
areas that need to be worked on (cf. e.g. Naumann, 2008, p.
139; Thomé & Thomé, 2014). The determinaon and the
assessment of these categories vary and are highly depending
on the purpose. The applied systems range from unpublished
templates developed by teachers to published and buyable
ones. For example, sciencally based and evaluated templates
for texts can be found in the “Oldenburger Fehleranalyse”
(OLFA) (Thomé & Thomé, 2014) and those specically meant
for the qualitave analysis of standardized tests can be found in
the “Aachener Förderdiagnossche Rechtschreibfehleranalyse”
(AFRA) (Herné & Naumann, 2002). When using the OLFA (Thomé
& Thomé 2014) the teacher has to collect texts containing
a certain amount of mistakes. Each mistake of a word has to
be analysed and categorized in one of the 35 categories (e.g.
lower case leer instead of upper case leer, upper case leer
instead of lower case leer, omission of a vowel, addion of
a vowel, etc.) that are described in detail in an accompanying
manual. Since the dierent categories are related to the stages
of spelling development the teacher gets to know the level of
spelling competence of a student.
In all described cases above, the me consuming analysis of
misspellings must be done by the teacher personally. This
requires eort to get familiar with the theory of German
orthography and the (applied) way of analysing the mistakes.
From our experiences, a detailed analysis is made only by highly
specialised people in rare cases. Since a clear qualitave analysis
of misspellings is the basis for a good and target oriented
intervenon, the IDERBLOG-Project aims to conduct the
analysis in large part automacally in order to support teachers
and consequently foster the spelling acquision process for
children.
Requirements for analysis
The categories of the qualitave analysis for the intelligent
diconary need to full some requirements on scienc,
technical and praccal basis. In order to full all these
requirements, the system of categories is established on
dierent hierarchical levels from ne to coarse grained. This
has the advantage, that the system stays exible as each level
is mainly dedicated to a specic purpose. We had to take into
account that many dierent words belong to one category
of mistakes. In order to provide a detailed analysis, we split
a category into specic phenomena (see table 1). Based on
those we have a proper ne-grained level for the applicaon
of learning analycs. However, those phenomena on such a
detailed level are not suitable for a general feedback. Therefore,
the phenomena of this specic level are merged in order to
retrieve a qualitave analysis for the teacher with a manageable
amount of categories and in order to be linked to the database
containing appropriate orthographic exercises. It also needs to
be taken into account that the naming of the categories that
are visible for the teachers and/or children, are sciencally
correct but sll easy to understand and consistent with the
established terms used in school environment, which are not
always consistent with the scienc terminology.
Method
In order to establish the dierent categories, a literature survey
was conducted and well-known approaches for qualitave
analysis for misspellings within the German orthography
were evaluated (cf. Edtstadler, in press). At the same me, 55
short texts of 3rd grade students from Germany and a limited
9
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
number of longer texts of 5th and 7th graders from Austria were
collected. In a next step a dra of categories was developed
based on the ndings of our research of relevant literature as
well as by using, as a starng point, established categories of
exisng analysis methods such as OLFA and AFRA. This dra
was rearranged, modied and extended in order to meet the
requirements of the intelligent diconary. It was especially
challenging to construct the categories in a way that the
descripon of the phenomena ts in with the possibilies of
programming misspellings as well as with the categories for
the teachers and the database with the exercises. Addionally,
specic misspellings due to the existence of dierent German
variees are considered.
The usability and suitability of the dra’s categories were further
checked by assigning one mistake encountered in the above
menoned texts from dierent regions of the German speaking
area to a phenomenon, for which a feedback can be given.
Also, phenomena of mistakes that were not found in the quite
small amount of texts’ (menoned above), but are theorecally
possible and/or by experience encountered in students’ texts
were added: For example, in the texts an inected form shows
that the ‘Umlaut’ is substuted by <e> (<*fengt> he/she/it
catches instead of <fängt> because of <fangen> to catch),
therefore, the substuon in plural forms (<Apfel - Äpfel>
apple - apples), derivaons (<Glanz – glänzend> shine - shiny),
and comparisons (<warm - wärmer> warm -warmer) were also
added.
General Description of the Categories
The categories are established on a linguisc and orthographic
basis, also by regarding previous ndings of the theory of German
Orthography (e.g. Nerius, 2007). Consequently, the system (see
table 1) is divided in two parts: On the one hand, the system
contains the parts that are invisible for the user where the
scienc and theorecal basis can be found. This is necessary
for the descripon of the phenomena and consequently for
programming the possibly misspelled words. The visible parts,
on the other hand, appear in the qualitave analysis for the
teacher, serve for the selecon of exercises from the training
database and appear in the feedback the writer gets in case a
word is not spelled correctly.
Since this system is quite complex, the described system is
shown in table 1. The orthographic area of ‘Umlaut’ serves as
an example.
Linguistic level
(not visible)
Ortho-graphic
area
(not visible)
description/
rule based
phenomenon
(not visible)
Category for
the teacher
(visible)
Category of
spelling
exercise
(visible)
Sub-category of
spelling
exercise
(visible)
Example of a
misspelled word
Feedback for
the writer
Morpho-logical
level
Um-laut Inflection of
nouns: e/eu for
ä/äu
Umlaut
derivable
Morpho-logical
hints
Derivation apples: not
<*Epfel> but
<Äpfel> because
of <Apfel> apple
Think, if there
exists a base
form with a.
Morpho-logical
level
Um-laut Inflection of
verbs: e/eu for
ä/äu
Umlaut
derivable
Morpho-logical
hints
Derivation he/she/it catches
er/sie/es not
<*fengt> but
<fängt> because
of <fang-en> to
catch
Think, if there
exists a base
form with a.
Morpho-logical
level
Um-laut Comparison of
adjectives: e/eu
for ä/äu
Umlaut
derivable
Morpho-logical
hints
Derivation warmer: not
<*wermer>
but <wärmer>
because of
<warm> warm
Think, if there
exists a base
form with a.
Morpho-logical
level
Um-laut Word formation
/ derivation: e/
eu for ä/äu
Umlaut
derivable
Morpho-logical
hints
Derivation shiny: not
<*glenzend>
but <glänzend>
because of
<Glanz> shine
Think, if there
exists a base
form with a.
Table 1: Example of the system of the intelligent diconary in the orthographic area of ‘Umlaut’
10
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
Starng from four linguisc levels – phonological,
morphological, lexical and syntacc – the categories are further
divided into orthographic areas. The phonological, the lexical
and the syntacc level consist of three orthographic areas.
The morphological level that is used for giving insight to the
complex system (cf. partly Nerius, 2007, p. 158) contains ve
orthographic areas, including ‘Umlaut’ (see table 1), which are
essenal for spelling words correctly:
1. Morpheme constancy: lexical and grammacal morphemes
are spelled the same way in compounding, derivaon and
inecon, even when the sound cannot be heard (e.g.,
<Ohrring> earring, <Weihnachten> Christmas because of
<weihen> hallow, <stehen> not <*stehn> because of the
sux –en, <Verkäufer> seller because of the prex ver-).
2. Morphological hints for using capital leers: the use of
capital leers for nouns is quite dicult and depending
on the syntax, but because of certain suxes, derivaons
can easily be idened as nouns that must be wrien with
capital leers, e.g. <*belohnung> gracaon because of
the sux –ung.
3. Morphological hints for not using capital leers: there
are also some suxes that indicate that a given word is
not wrien with a capital leer, although morpho-syntax
can change the word class, e.g. <*Furchtbar> horrible is
correctly spelled <furchtbar> because of the sux –bar,
but it is spelled with capital leers in the phrase <etwas
Furchtbares> something horrible whereas the use of
the same word form as an adjecve requires the use of
lower case, e.g. <ein furchtbares Gewier> a horrible
thunderstorm.
4. ‘Umlaut’: Because of phoneme-grapheme correspondences,
especially in the area of Austria, the ‘Umlaut’ is oen
incorrectly wrien as <e>, e.g. <*glenzend> instead of
<glänzend> shiny, since the ‘Umlaut’ needs to be applied
because of the base morpheme <Glanz> shine (for details,
see table 1).
5. Terminal devoicing: In German a word is pronounced
with a devoiced obstruent at the end of the word, but
spelled with the voiced variant of the phoneme-grapheme
correspondences (e.g. /hunt/, but spelled as <Hund> dog
because of the wordform /hunde/ whereas <Brot> bread is
spelled as <Brot> because of /bro:t – bro:tə/).
Each orthographic area is associated with a wide range of
phenomena. These phenomena are formed in a way that they
can funcon as a rule for programming the possible mistakes
(see table 1). The number of phenomena is depending on the
given orthographic area and can be expanded and reduced,
based on evidence. The following example will help to show the
variety in the number and the characteriscs of phenomena:
In the orthographic area of morpheme constancy the
category derivaonal suxes (for the analysis for the teacher)
summarises the phenomenon of misspelling suxes such as –ig
(e.g. <lusg> funny) (wrien in dierent ways depending on the
spoken German variety as –ich <*lusch>, -isch <*lussch>, -ik
<*lusk>), and the phenomenon of spelling the sux –lich as
<*-lig> as well as further phenomena describing the misspelling
of other derivaonal suxes.
As menoned above, it is important to work with a manageable
amount of categories when oering the qualitave analysis for
the teacher. Therefore, the currently 110 phenomena are linked
with 34 categories of the qualitave analysis for the teacher.
In the example in table 1, the four dened phenomena for
misspelling the ‘Umlaut’ are summed up in one category that
tells the teacher that within a certain amount of mistakes the
‘Umlaut’ was derivable, but incorrectly spelled with the wrong
grapheme.
The categories of the qualitave analysis are then connected
with and/or mirrored in the labels of the orthographic exercises
available on the plaorm. For an easier orientaon they are
divided in categories and sub-categories of exercises. However,
the labelling of the exercises is in some cases more coarsely
grained than the category of the qualitave analysis itself (e.g.
the category upper case instead of lower case and the category
lower case instead of upper case are labelled as upper and
lower case exercises, since a lot of exercises pracce both at
the same me). This is due to the fact, that in the rst step only
already exisng exercises are available on the plaorm, but in
the progress of the project specic exercises will be developed.
The phenomena also funcon as the starng point for the
feedback of the intelligent diconary, which will be described
more in detail below.
5. Feedback from the Intelligent Dictionary
All of the categories and phenomena form the basis for the
analysis and applicaon of the intelligent diconary, which is
the core of the plaorm. The idea is that a child, who misspells
11
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
a word, does not only get the feedback that the word is spelled
incorrectly, but also gets a hint for correcng it. The given
feedback is connected with the phenomenon. In order to keep
a straighorward number of feedbacks the same feedback will
be given – whenever possible - for more than one phenomenon
within an orthographic area. It is formulated in a way that forces
the child to think about the spelling and further encourages
the development and applicaon of spelling strategies.
Therefore, no direct commands (e.g. “use <ä> instead of <e>”)
for correcng the word are included in the feedbacks. The
correcon will only be successful if the child reects on the hint
in combinaon with the misspelled word. This approach stands
in contrast to the usual word-correcon process where either
the misspelled word is marked or the correct word needs to
be selected from a variety of oered words. In both cases the
correcon will probably not lead to a deeper understanding of
correct orthography.
Klicpera et al. (2003, p. 255) menon that in order to acquire
correct spelling, it is important to oer exercises that allow the
autonomous correcon in a movang context. Experienced
teachers and trainers for dyslexic children know that poor
spellers have problems in idenfying their mistakes in a text.
But as soon as a hint for correcng the word is given, they oen
know how to spell it correctly. This is a successful, but a me and
energy consuming way of improving orthographic competence.
In order to avoid exhauson, the intelligent diconary gives this
feedback instead of a teacher, which also has the advantage
that the intelligent diconary can and will repeat the feedback
several mes. In case a child spells <*Epfel> instead of <*Äpfel>
apples the intelligent diconary provides the feedback “Think
if there exists a base form with “a”?”, the same feedback will
be given in case the child writes <*fengt> instead of <fängt> to
catch or <*glenzend> instead of <glänzend> shiny. There is no
strict applicaon of a syntacc paern for the feedback since
the wording of the feedback is chosen rather on a didacc than
on a formal basis.
Wordlist of the intelligent dictionary
Since this intelligent diconary so far, is designated to funcon
as a rst prototype, only a selecon of words funcons as the
basis for programming the diconary. For the rst prototype
we had to choose around 1000 words. Generally such a
selecon of words would be based on the frequency of the
CELEX (1995) database – although this would propose some
problems (cf. Brysbaert et al., 2011). Selecng words only based
on frequency in general without considering the frequency of
words in children´s language is especially problemac for the
development of an applicaon aimed at children. Also the fact
that the selected words should be prone to be misspelled had to
be considered (for a discussion see Risel, 2008).
In order to meet these requirements, the word list for the
prototype of the intelligent diconary is based on the basic
vocabulary of three German Federal States (Bavaria, Hamburg,
Berlin-Brandenburg). In the next step it was checked, whether
in these basic vocabularies the 100 most frequently misspelled
words of 4th graders (compiled and made available by Tacke,
2008) are included. Words that do not appear in any form in one
of these three basic vocabularies were included (e.g. <kommt>
comes was not included since <kommen> to come is a word of
the basic vocabulary, but <ziemlich> quite was included since
it does not appear in one of the basic vocabularies). At the
end, the word list for the prototype of the intelligent diconary
ended up containing around 1100 words.
Since German has a rich morphology and texts are not merely
made of words that are listed in a diconary, it is necessary
to list all the possible word forms of a given word in order to
construct all possible misspellings for the intelligent diconary
in a next step. The collecon of all possible word forms of a
given word (or to be precise of a lemma) is based on the CELEX
(1995) database. This incorporaon of all word forms enlarged
the wordlist to over 7500 orthographically correct words. In
German the variaon in the number of word forms for a given
lemma is quite high as is proven by the following examples:
• For the adjecve <ähnlich> similar the CELEX (1995)
database has 17 word forms (<ähnlich, ähnliche, ähnlichen,
ähnlicher, ähnlichem, ähnliches, ähnlichst, ähnlichste,
ähnlichsten, ähnlichster, ähnlichstem, ähnlichstes,
ähnlichere, ähnlicheren, ähnlicherer, ähnlicherem,
ähnlicheres>) including inecon for singular, plural, the
dierent cases and comparison.
• For the regular verb <arbeiten> to work 10 word forms
can be found in the CELEX (1995) database in summary,
whereas for the irregular verb <beginnen> to begin exactly
24.
• For the noun <Beispiel> example exist only four word forms.
• For preposions no other word form can be found since
they cannot be modied.
12
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
For these 7500 word forms of this wordlist all possible mistakes
for a given category are constructed and connected with the
feedback. For instance, for the orthographic area of ‘Umlaut’
all words containing an <ä> in the wordlist must be searched
for. To consider the dierent phenomena where substung
<ä> for <e> is a mistake due to the morphological principle,
the search is done separately for verbs (e.g. <*fengt> instead
of <fängt> he/she/it catches), nouns (e.g. <Epfel> instead of
<Äpfel> apples), adjecves (e.g. <*kelter> instead of <kälter>
colder) and derivaons (e.g. <glenzend> instead of <glänzend>
shiny). But, since the ‘Umlaut’ in the word <ähnlich> and its
substuon with the incorrect <e> does not qualify for the
morphological level it must not be included in this category but
has to be added in another appropriate category with according
feedback.
6. Conclusion
In this arcle we introduced a plaorm that aims to movate
children to improve their spelling skills by wring and publishing
texts. In this plaorm an intelligent diconary is integrated and
based on the presented system of categories, the intelligent
diconary gives feedback in order to enable children to
correct mistakes with the help of this feedback. The plaorm
also provides a qualitave analysis for teachers, who can use
the results in order to help pupils with the improvement of
word spelling. Concerning learning analycs, the occurred
misspellings can also be used for an in depth analysis.
The development of the plaorm and the intelligent diconary
is sll under construcon and changes are sll possible. There
are sll issues such as the idencaon of several mistakes in
one word that will most certainly lead to further discussion
in the future. However, we are posive that this combinaon
and the interdisciplinary work of the IDERBLOG-Project will in
future movate more children from grade 3 on to write texts
and to improve their spelling competence. Further, we can
support teachers by providing analysis and material for the
improvement of spelling. The acve applicaon of the methods
of learning analycs in this area of language learning will help us
to understand the process of spelling acquision in more detail.
It is expected that this unique combinaon in one plaorm has a
posive impact on didacc approaches, educaon and science.
7. Acknowledgements
This research project is supported by the European Commission
Erasmus+ program in the framework of the project IDERBLOG.
For more informaon about the IDERBLOG-Project and its
project partners: Hugo Adolph2 , Chrisan Aspalter3 , Susanne
Biermeier4 , Sandra Ernst5 , Sonja Gabriel 6, Gabriele Goor5
Michael Gros2, Mike Cormann5, Anneliese Huppertz5, Kathrin
Irma4, Susanne March3, Nina Steinhauer2, Behnam Taraghi
7 and Marianne Ullmann3, please visit our homepage hp://
iderblog.eu/ (German language only).
2 LPM Saarland, Beethovenstraße 26, 66125 Saarbrücken, Germany
3 University of Teacher Educaon Vienna, IBS/DiZeTIK, Grenzackerstraße 18,
1100 Vienna, Austria - Europe
4 Albert-Weisgerber School St. Ingbert, Robert-Koch-Straße 4, 66386 St. Ingbert,
Germany
5 School of Raeren, Hauptstraße 45, 4730 Raeren, Belgium
6 University College of Teacher Educaon Vienna/Krems, Mayerweckstraße 1,
1210 Vienna, Austria - Europe
7 Graz University of Technology, Department Educaonal Technology,
Münzgrabenstraße 35a, 8010 Graz, Austria - Europe
13
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
References
Baker, R. S. J. D., Duval, E., Stamper, J., Wiley, D. & Buckingham Shum,
S. (2012). Panel: educaonal data mining meets learning analycs. In
Buckingham Shum, S., Gasevic, D. & Ferguson, R. (Eds.) Proceedings of 2nd
Internaonal Conference on Learning Analycs & Knowledge (LAK’12),
New York, USA, 20.
Bartnitzky, H. (2010). Individuell fördern–Kompetenzen stärken.
Grundschule aktuell (9), 6-11.
Broc, L., Bernicot, J., Olive, T., Favart, M., Reilly, J., Quémart, P. &
Uzé, J. (2013). Lexical spelling in children and adolescents with specic
language impairment: Variaons with the wring situaon. Research in
Developmental Disabilies 34, 3253–3266.
Brysbaert, M., Buchmeier, M., Conrad, M., Jacobs, A.M., Bölte, J. & Böhl,
A. (2011). The word frequency eect: a review of recent developments
and implicaons for the choice of frequency esmates in German. In:
Experimental Psychology 58(5), 412-424.
Campbell, J. P., DeBlois, P. B. & Oblinger, D. G. (2007). Academic analycs:
A new tool for a new era. EDUCAUSE review, 42(4), 40.
CELEX (Center for Lexical Informaon) - Linguisc Data Consorum
(1995). The CELEX Lexical Database, Release 2, German, Version 2.5.
Clow, D. (2012). The learning analycs cycle: closing the loop eecvely. In
Buckingham Shum, S., Gasevic, D. & Ferguson, R. (Eds.) Proceedings of 2nd
Internaonal Conference on Learning Analycs & Knowledge (LAK’12),
New York, USA, 134-138.
Duval, E. (2010). Aenon Please! Learning Analycs for Visualizaon and
Recommendaon. In Proceedings of LAK11: 1st Internaonal Conference
on Learning Analycs and Knowledge 2011, Retrieved October 8, 2015
from hps://lirias.kuleuven.be/bitstream/123456789/315113/1/la2.pdf
Ebner, M., Taraghi, B., Ebner, M., Aspalter, C., Biermeier, S., Edtstadler,
K., Gabriel, S., Goor, G., Gros, M., Huppertz, A., March, S., Steinhauer,
N., Ullmann, M. & Ziegler, K. (2015a). Design für eine Plaorm zum
Schreibenlernen im Grundschulalter. In Rathmayer, S. & Pongratz, H. (Eds.)
Proceedings of DeLFI Workshops 2015 co-located with 13th e-Learning
Conference of the German Computer Society (DeLFI 2015) Munich,
Germany, September 1st, 2015, 118-122.
Ebner, M., Taraghi, B., Saran, A. & Schön, S. (2015b). Seven features of
smart learning analycs-lessons learned from four years of research with
learning analycs. eLearning papers 40, 51–55.
Edtstadler, K. (in press). Qualitave Fehleranalyse im Schrispracherwerb.
Krik und Kriterien. In Lindner D., Gabriel S. & Beer R. (Eds.) Dialog
Forschung – Tag der Forschung 2015, Lit-Verlag.
Fuhrhop, N. (2011). System der Getrennt- und Zusammenschreibung.
In Bredel, U. & Reißig, T. (Eds.) Weiterführender Orthographieerwerb.
Schneider Verlag Hohengehren, 107-128.
Flor, M. & Futagi, Y. (2012). On using context for automac correcon of
non-word misspellings in student essays. In Proceedings of the Seventh
Workshop on Innovave Use of NLP for Building Educaonal Applicaons,
Associaon for Computaonal Linguiscs, 105-115.
Fröhler, H. (2002). Fernitzer Grundwortschatz 2002. Wortliste für die
Grundschule. Ausgabe 2002. Wien: Eigenverlag.
Greller, W. & Drachsler, H. (2012). Translang learning into numbers:
A generic framework for learning analycs. Journal of Educaonal
Technology & Society, 15(3), 42-57.
Government of South Australia. Department for Educaon and Child
Development, issuing body. (2011). Spelling: from beginnings to
prociency: a spelling resource for planning, teaching, assessing and
reporng on progress, [Adelaide, South Australia]: Department for
Educaon and Child Development. Retrieved October 8, 2015 from
http://www.decd.sa.gov.au/literacy/files/pages/Programs%20and%20
Resources/Spelling_resource_FINAL.pdf
Herné, K.-L. & Naumann, C.L. (2002). Aachener Förderdiagnossche
Rechtschreibfehler-Analyse, Alfa-Zentaurus.
Holzinger, A., Errath, M., Searle, G., Thurnher, B. & Slany, W. (2005). From
extreme programming and usability engineering to extreme usability in
soware engineering educaon (XP+ UE→ XU). In Computer Soware
and Applicaons Conference, 2005. COMPSAC 2005. 29th Annual
Internaonal. (2), IEEE, 169-172.
Høien, T. & Lundberg, I. (2000). Dyslexia: From Theory to Intervenon.
Kluwer.
Katz, L. & Frost, R. (1992). The reading process is dierent for dierent
orthographies: The orthographic depth hypothesis. Haskins Laboratories
Status Report on Speech Research, 147-160. Retrieved October 8, 2015
from hp://www.haskins.yale.edu/sr/SR111/SR111_11.pdf
Khalil, M. & Ebner, M. (2015). Learning Analycs: Principles and
Constraints. In Proceedings of World Conference on Educaonal
Mulmedia, Hypermedia and Telecommunicaons, EdMedia 2015. AACE
Waynesville, NC, USA, 1326-1336.
14
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
Klees, M. (1989). Developmental Dyslexia in French. In Aaron, P.G. &
Malatesha Joshi, R. (Eds.) Reading and Wring Disorders in Dierent
Orthographic Systems. Kluwer Academic Publishers, 137-142.
Klicpera, Ch., Schabmann A. & Gasteiger-Klicpera, B. (2003). Legasthenie.
Reinhardt.
Küel, H. (2007). Probleme des Erwerbs der Orthographie. In Nerius, D.
(Ed.) Deutsche Orthographie. Georg Olms, 417-451.
Landerl, K. & Thaler, V. (2006). Reading and spelling acquision and
dyslexia in German. In Joshi, R.M. & Aaron, P.G. (Eds.) Handbook of
Orthography and Literacy, Lawrence Erlbaum, 121-134.
Liebal, J. & Exner, M. (2011). Usability für Kids. Springer Fachmedien
Wiesbaden.
May, P. (2010). Hamburger Schreib-Probe 1-9. Manual. Kle.
Meinhold, G. & Stock, E. (2007). Phonem-Graphem-Beziehungen in
deutschen Wörtern. In Nerius, D. (Ed.) Deutsche Orthographie. Georg
Olms, 108-128.
Naumann, C. L. (2008). Zur Rechtschreibkompetenz und ihrer Entwicklung.
In Bremerich-Vos, A., Granzer, D. & Köller, O. (Eds.) Lernstandbesmmung
im Fach Deutsch, Beltz, 134-159.
Nerius, D. (2007). Deutsche Orthographie, Georg Olms.
Neuhold, B. (2013). Learning Analycs-Mathemak Lernen neu gedacht,
BoD–Books on Demand.
Piety, P. J. (2013). Assessing the educaonal data movement, Teachers
College Press.
Risel, H. (2008). Arbeitsbuch Rechtschreibdidakk, Schneider Verlag
Hohengehren.
Schön, M., Ebner, M. & Kothmeier, G. (2012). It’s just about learning the
mulplicaon table. In Proceedings of the 2nd Internaonal Conference
on Learning Analycs and Knowledge, ACM, 73-81.
Siemens, G. (2012). Learning analycs: envisioning a research discipline
and a domain of pracce. In Proceedings of the 2nd Internaonal
Conference on Learning Analycs and Knowledge, ACM, 4-8.
Siemens, G. & Long, P. (2011). Penetrang the Fog: Analycs in Learning
and Educaon. EDUCAUSE review, 46(5), 30.
Snowling, M. J. (1989). Developmental Dyslexia: A Cognive
Developmental Perspecve. In Aaron, P.G. & Joshi, R.M. (Eds.) Reading
and Wring Disorders in Dierent Orthographic Systems, Kluwer Academic
Publishers, 1-23.
Sommer Stumpenhorst, N. (2012). Rechtschreibkompetenz entwickeln.
Retrieved October 9, 2015 from hp://www.rechtschreib-werksta.de/rsl/
gk/html/kompetenz.html
Tacke, G. (2008). Rechtschreibüben als Projektunterricht für die
Klassenstufen 4 bis 8. Retrieved October 9, 2015 from hp://www.
cyberkidz.de/pdf/Projektunterricht-worter-5-7.pdf
Thelen, T. 2010. Automasche Analyse orthographischer Fehler von
Schreibanfängern. PhD-thesis: University of Osnabrück. Retrieved
November 8, 2015 from hps://repositorium.uni-osnabrueck.de/
bitstream/urn:nbn:de:gbv:700-201006096307/1/thesis_thelen.pdf
Thomé, G. & Thomé, D. (2014). Oldenburger Fehleranalyse für die Klassen
3-9, Instrument und Handbuch, Instut für sprachliche Bildung.
Taraghi, B., Saran, A., Ebner, M., Müller, V. & Großmann, A. (2015a).
Towards a Learning-Aware Applicaon Guided by Hierarchical Classicaon
of Learner Proles, Journal of Universal Computer Science, 21(1), 93-109.
Taraghi, B., Frey, M., Saran, A., Ebner, M., Müller, V. & Großmann, A.
(2015b). Determining the Causing Factors of Errors for Mulplicaon
Problems. In Ebner, M., Erenli, K., Malaka, R., Pirker, J., Walsh, A. (Eds.)
Immersive Educaon, Springer, 27-38.
Tsesmeli, S. N. & Seymour, P.H.K. (2006). Derivaonal morphology and
spelling in dyslexia. Reading and Wring, 19(6), 587–625.
Valn, R. (1989). Dyslexia in the German Language. In Aaron, P.G. &
Joshi, R.M. (Eds.) Reading and Wring Disorders in Dierent Orthographic
Systems, Kluwer Academic Publishers, 119-135.
15
In-depth
eLearning
Papers
45
eLearning Papers • ISSN: 1887-1542 • www.openeducationeuropa.eu/en/elearning_papers
n.º 45 • November 2015
Copyrights
The texts published in this journal, unless otherwise indicated, are subject
to a Creative Commons Attribution-Noncommercial-NoDerivativeWorks
3.0 Unported licence. They may be copied, distributed and broadcast pro-
vided that the author and the e-journal that publishes them, eLearning
Papers, are cited. Commercial use and derivative works are not permit-
ted. The full licence can be consulted on http://creativecommons.org/
licenses/by-nc-nd/3.0/
Edition and production
Name of the publication: eLearning Papers
ISSN: 1887-1542
Publisher: openeducation.eu
Edited by: P.A.U. Education, S.L.
Postal address: c/Muntaner 262, 3r, 08021 Barcelona (Spain)
Phone: +34 933 670 400
Email: editorialteam[at]openeducationeuropa[dot]eu
Internet: www.openeducationeuropa.eu/en/elearning_papers
