Improved Learning of Academic Writing

Reducing Complexity by Modeling Academic Texts

Gert Faustmann

Department of Cooperative Studies, Hochschule f

ur Wirtschaft und Recht Berlin (Berlin School for Economics and Law),

Alt-Friedrichsfelde 60, 10315 Berlin, Germany

Keywords:

Text Coherence, Writing Process, Argument Representation, Text Perspectives, Text Patterns.

Abstract:

A graphical modeling language for scientiﬁc texts is presented, which particularly supports the learning pro-

cess of academic writing. The supervision process for textual work in higher education is often characterized

by misunderstandings, since the agreements are based on the abstract level of the document outline. The writ-

ten text then often misses an inner structure and suitable representations and thus has little coherence. Since

an academic text is a complex combination of text segments, linguistic functions, content, and means of pre-

sentation, a notation is proposed that includes these different perspectives. Based on the UML for software

modeling, extracts of a text as well as different levels of abstraction can so be part of the learning process.

Finally, a software tool is sketched, which can support the construction of document frameworks as well as

the creation of the actual text.

1 INTRODUCTION

In scientiﬁc discourse, publications are a fundamental

part of communication. The quality of scientiﬁc texts

is not only determined by the quality of the research

work carried out, but also by the clarity of the presen-

tation of the results obtained. This property of texts

is called coherence (Beaugrande and Dressler, 1981)

and expresses itself in scientiﬁc texts as an inner con-

nection of argumentations and the parts contained the-

rein (statement, evidence, justiﬁcation, qualiﬁcation)

(Booth et al., 2003).

Academic papers at universities are examinations

in all phases of the study programme. As the course

progresses, the complexity of the texts to be produced

increases. The creation process is not only inﬂuenced

by the student, but also by an academic supervisor

during the planning process. However, misunderstan-

dings often arise here, as the subject of the discussion

is more likely to be at the abstract level of the docu-

ment structure (which means the outline with chap-

ters and subsections). Concrete content-related cha-

racteristics of the individual sections in the text only

become clear after submission of the work and often

exhibit a lack of coherence.

There is therefore a gap between the planning of

a scientiﬁc text at its structure level (document struc-

ture) and its content characteristics (functional struc-

ture). A well-arranged representation of this functio-

nal structure in between could on the one hand make

the support process in teaching more efﬁcient and on

the other hand fundamentally improve text production

in the academic ﬁeld.

2 SUPPORT FOR ACADEMIC

WRITING PROCESSES

The necessity of supporting writing processes is re-

cognised in university education and is also imple-

mented by various measures. Considering the cur-

rent situation at universities, there are common appro-

aches to support academic writing. However, various

authors also suggest more extensive means to improve

the process of writing and supervising a thesis. One

of these is stronger planning and modeling.

2.1 Common Approaches

In order to promote the writing process at universities

and colleges, three approaches can be distinguished:

• In courses on scientiﬁc work and academic wri-

ting, behaviors are taught to plan a written work,

to collect information and to ﬁnd a question

whose solution with suitable representations re-

sults in a text. In this context, textbooks dealing

with academic writing are also used.

Faustmann, G.

Improved Learning of Academic Writing.

DOI: 10.5220/0006792204470453

In Proceedings of the 10th International Conference on Computer Supported Education (CSEDU 2018), pages 447-453

ISBN: 978-989-758-291-2

447

• Through examples of dissertations (e. g. from the

university library) students can see how concrete

topics were worked on and speciﬁc problems were

solved in detail (e.g (Cofﬁn et al., 2002) descri-

bing an argument essay outline).

• In speciﬁc support processes of written disserta-

tions research questions are chosen and formula-

ted, approaches and structures are discussed, as

well as formal criteria are clariﬁed. Text excerpts,

graphic representations, structures and abstracts

are used for communication.

If there is a writing task in the course of studies, stu-

dents are confronted with various problems:

• On the one hand, the contents of the courses on

scientiﬁc writing date back a long time and are no

longer directly available as practical knowledge.

In addition, most of the contents were not focused

enough on the current question of the work to be

done.

• The sample papers provide an insight into the de-

sign and framework conditions of a ﬁnal thesis,

but they usually cannot be mapped to the que-

stion, method or argumentation of the student’s

own task.

• Due to the formal framework conditions, such as

limited time for discussion and only rough des-

criptions of the text to be written, the support pro-

cess remains at a rather abstract level.

2.2 External Document Representations

In general, the use of external representations of

thought processes on different levels is recognized as

the basis for a qualitatively better design performance

(Kirsh, 2010).

Various forms of externalisation have found their

way into scientiﬁc text production in recent years.

Mindmaps are generally known (Buzan, 2006) and

serve the hierarchical organization of questions. They

can also be easily combined with brainstorming pro-

cesses. In this context, concept maps are also known,

which do not have to be hierarchically structured and

combine ideas (concepts) in a network structure (e. g.

(Heard, 2016)).

The general explication of argumentation was al-

ready described in early studies (Newell, 1979) (Dijk,

1980) (Toulmin, 2003). Some authors design speciﬁc

models for argumentation chains in e-learning ((Bell,

1997) (Faustmann, 2011) and in hypertext systems

(Neuwirth and Kaufer, 1989) (Streitz et al., 1992).

Hausdorf examines, for example, the entire deve-

lopment process of a scientiﬁc work with informa-

tion objects involved in it (Hausdorf, 2005). The text

structure is also discussed here and in the implemen-

ted software tool ScientiFix hierarchical representati-

ons of sections, ideas and sources can be found (e. g.

Figure 7.12, p. 193). In (Byrne and Tangney, 2010) a

tool is developed that includes different representati-

ons of a document (map view, tree view and text view)

based on the framework of Shibata and Hori (Shibata

and Hori, 2002) for organizing document parts.

The perspectives on texts used in Shibata and Hori

are strongly reminiscent of outlines (hierarchical per-

spective) and concept maps (map perspective). Both

are nowadays still well known means to support the

writing of texts. The overall problem of external re-

presentations of texts is, that there is no integrated

model for planning the document and text structures

on all levels that make up an academic text.

3 CONCEPT OF A TEXT

MODELING LANGUAGE

The various problems, which are frequently encoun-

tered in more extensive written work, show that stu-

dents have to cope with the complexity of the functi-

onal parts of the text and their interrelationships. For

this reason, we ﬁrst want to analyse brieﬂy how the

complexity of other systems to be constructed can be

mastered.

3.1 Modeling Complex Products

The problem of complexity can be similarly encoun-

tered in the construction of software artifacts: a com-

plex software system can no longer be overlooked in

its many tasks and arising dependencies. In recent

decades, various paradigms have been established to

structure a system (e. g. object orientation) and to

plan clearly with suitable representation methods (e.

g. Uniﬁed Modeling Language UML).

In the design, two types of presentation can be dis-

tinguished in principle: on the one hand, the logical

design, which describes the task of the system and

the subject-speciﬁc solution to the problem, and on

the other hand, the system design, which is to repre-

sent an exact model of the later system. The advan-

tage here lies in the separation of the algorithmic lo-

gic from technical decisions: both interact with each

other, but should initially be unaffected by each ot-

her. The UML not only offers the possibility to dis-

play both design layers, but also to convert them into

each other without any problems. For example, the

Domain Driven Design methodology implements this

approach in a software process model (Evans, 2003).

CSEDU 2018 - 10th International Conference on Computer Supported Education

448

In addition, different views of the architecture of

a software system can be described in UML: typical

examples are the structural and behavioral view. In

this way, class diagrams can describe the basic struc-

ture of a system and component diagrams can depict

the division of classes into modules. Other diagrams

exist, for example, for offered functions (use case di-

agrams), the physical distribution to host nodes (de-

ployment diagrams) and communication between the

objects of the system (sequence diagrams) (see Figure

1).

Structure of a System

Class Diagram

Communication

in a System

Sequence diagram

Physical

Distribution

of a System

Deployment diagram

Functions of

a System

Use Case Diagram

Divison in!

Subsystems

Component diagram

Figure 1: Perspectives on Software in UML.

These different levels of abstraction thus lend them-

selves to being transferred to text production.

3.2 Requirements for Academic Text

Modeling

3.2.1 Ease of Use

To reduce the complexity of a product, it is neces-

sary to reduce the amount of notational elements of

a system to a minimum. This does not mean that the

models will only have a small scope, but the learning

effort for using the language elements will be consi-

derably reduced. Thus the ﬁrst large requirement area

of a text modeling language is simple use. This inclu-

des being able to grasp a ﬁrst model relatively quickly

and, if necessary, to sketch a model by hand.

3.2.2 Different Perspectives

The next requirement concerns visibility of different

perspectives on the text to be made. Well-known per-

spectives are the text structure with its outline, the re-

presentations within texts by continuous text, ﬁgures,

tables, etc., textually described parts of the later text,

as well as functions of the text components such as

analyses, hypotheses, summaries, etc.

3.2.3 Relationships

Furthermore, it must be possible to create relations-

hips between text elements, both within and across

perspectives. Examples of relationships are the local

sequence in the text (e.g. chapter 2 follows chapter

1), the referencing of parts of the text that are contai-

ned earlier in the text and will only appear later (e.g.

the comparison of an examination of a foreign author

with own results) and the inclusion of a part of the text

by a function (e.g. the above-mentioned comparison

can be an evidence for a certain hypothesis).

3.3 Language Description

The proposed language for modeling academic texts

contains various notation elements, each of which

is assigned to one of four perspectives. An impor-

tant perspective is that of the functional parts, which

must meet the requirements of scientiﬁc proof. Often,

works also show shortcomings in the presentation of

the results, to which an own perspective is respon-

ding. Finally, relationships between the different text

elements indicate dependencies at all levels. See the

now detailed language elements in an overview model

in Figure 2.

3.3.1 Perspectives

The selection of the contained perspectives is based

on two basic conditions: on the one hand, in today’s

writing processes different means are already used to

structure and design texts (see Section 2.1). These

include structures, collections of ideas and concept

maps. On the other hand, the main problems lie in

the preparation of scientiﬁc papers in elaboration of

the argumentation, as well as in the appropriate pre-

sentation of the results (see introduction). This leads

to the following four perspectives:

1. Outline Perspective

This section describes the division of the text into

chapters, subchapters and sections. It becomes

clear which parts follow each other or contain ot-

her parts of the text.

2. Functional Perspective

Here, functional units are identiﬁed and their in-

terrelationships are recorded by appropriate rela-

tionships. From this perspective, it becomes clear

which descriptions and examples have led to a

hypothesis, for example, and which analyses sup-

port a hypothesis.

Improved Learning of Academic Writing

449

EVI

DOCUM

EXAMPL

1. Introduction

2. Implementation 3. Evaluation

3.1 Engagement 3.2 Attainment

4. Discussion

HYPOTH

Outline

Function

Content

Presentation

DOCUM

student

generated

content (sgc)!

provided

functions

PeerWise

correlation

to exam

outcomes

DOCUM

PeerWise

logging

data

DOCUM

student

evaluation

questonnaire

ANALYS

activity

based

median

split

DOCUM

mean

marks

COMP EVID

comparison

end course

mark with

PeerWise

activity level

evidence of

better

students

performing

more highly

ANALYS

median split

for ability

quartiles

COMP

even lowest

quartile

achieves

better

outcome

enhanced

conceptual

understanding

Figure 2: Example for a Text Model.

3. Content Perspective

In this perspective, short descriptions of the con-

tent are assigned to the text elements, which can

give the viewer a ﬁrst idea of the content. These

can be compared with concepts in concept or

mind maps.

4. Presentation Perspective

Text elements can be presented very differently.

Thus, an analysis of different methods can be

done in tabular form or in continuous text. Further

known types of presentation are ﬁgures, mathe-

matical representations (e.g. equations) and pro-

gram listings or algorithms.

3.3.2 Functional Elements

For scientiﬁc work, functional elements of a text

can be subdivided into the areas of knowledge

collection/preparation, knowledge creation and

knowledge classiﬁcation (Hausdorf, 2005).

1. The knowledge collection includes

Collection/Documentation. A selection of the

objects to be viewed (DOCUM) or an example

representation (EXAMPL) is made.

Analysis. How is an object broken down into its

components (ANALYS)?

Comparison. Which parts of an object are consi-

dered important (COMP)?

2. Knowledge creation is characterized especially by

reasoning which contains the following compo-

nents:

Hypothesis. What can be assumed on the basis

of previous considerations (HYPOTH)?

Evidence. What speaks for a hypothesis (EVID)?

Evaluation. How to assess this argumentation

(EVAL)?

3. At knowledge classiﬁcation we need:

Summary. What is the path of a hypothesis and

what are the consequences of the now proven

hypothesis (SUMM)?

Comparison. What relation does new know-

ledge have to the already existing knowledge

(COMP)?

The function elements listed here appear in the

function view as symbols with their respective abbre-

viations.

3.3.3 Presentational Elements

Scientiﬁc texts contain the following types of content

presentation:

• Text: a text without additional structuring measu-

res. Typically, a unit of text is grouped together as

a paragraph with one thought in it.

• Enumerations or Lists: if different parts of a text

of one type are described, they can be separated

from each other by dots (point / minus signs) and

presented in a structured way for the reader. Enu-

merations number the text parts.

• Tables: they are mostly used to display ﬁgures,

but can also contain symbolic information (e.g.

identiﬁers).

• Figures: are of very different types and serve to

illustrate connections.

• Program Code/ Algorithms: if procedures or even

concrete implementations are to be shown, this

can be done through suitable languages of an al-

gorithmic nature.

CSEDU 2018 - 10th International Conference on Computer Supported Education

450

• Mathematical Expressions: describe numerical

relationships.

Presentation elements are indicated in the text mo-

del by a pictorial representation. Figure 3 shows the

assignment of function elements to screen representa-

tions.

int x=0;!

x++;

Figure

Math

Program

Table

Text

List

Figure 3: Presentational Elements.

These presentation elements are partly interchangea-

ble. For example, the content of an illustration can

also be described in continuous text. On the other

hand, however, it makes sense in some places to de-

pict a content to be represented by more than one pre-

sentation element (e.g. the content of an illustration

should be explained additionally in the text).

3.3.4 Relationships

Relationships between text elements are of different

nature: the simplest connection is the sequence within

the text to be written. Thus chapters follow each ot-

her, but also functional units such as an example and

an analysis. If the sequence is not clear from the ar-

rangement in the model, a connecting line with an un-

ﬁlled arrowhead can be used.

Another relation is the provision or the inclusion.

Thus, chapters may include subchapters, but a compa-

rison can also provide evidence for a hypothesis. The

relationship is visualized by a simple connecting line.

After all, parts of the text must also be able to re-

late to each other in their statement. This can be inter-

preted as a reference that may not have any overlap-

ping content. In this way, a later derived evidence can

also be used for a hypothesis presented in the text.

The reference relationship must be made clear here.

This relationship can be shown by a line with an open

arrowhead (see Figure 4).

HYPOTH EVID

EXAMPL ANALYS

COMP

sequence

inclusion

reference

Figure 4: Relationships.

4 CONCLUSIONS

4.1 Advantages of Text Modeling

The approach described here to modeling academic

texts explains the argumentative background of a

work, as well as the means of presenting the neces-

sary content. This makes it possible to carry out a

more detailed planning of a text in terms of its signi-

ﬁcance before the actual writing process begins.

Furthermore, it is now also clear for existing texts

how they are structured without having to be read in

detail. It would even be conceivable to offer the es-

sence of a text, i.e. hypotheses of interest to the rea-

der, as well as their proofs by means of documented

evidence, in a compact form and to refer only to these

passages of the text.

Both possibilities could be used for future learning

processes: in the context of supervising dissertations,

the construction process of the text could be improved

and in the context of lectures, the analysis of existing

research reports on the basis of a text model would be

much easier to understand for students.

4.2 Limitations

However, the concept also presents difﬁculties: ﬁrst

of all, the language deﬁnition must be available to

and understood by all those involved. This involves

a certain amount of extra effort, which could be put

into perspective within the respective courses and by

means of suitable manuals.

One danger in the use of system notations is the

supposed security of getting to high-quality systems.

The use of engineering methods and notations such

as the Uniﬁed Modeling Language is by no means

a guarantee for the speciﬁcation of appropriate and

error-free software. However, these methods increase

the probability of software quality by improving the

overview.

Text models can be used to create a framework

for scientiﬁc texts that contains important parts, but

the texts can still be difﬁcult to understand, do not

correspond to the assigned functional units and much

more.

5 FUTURE WORK

In order to ensure that the concept of modeling acade-

mic texts also brings the desired beneﬁts in practice,

the following additional work is necessary:

Improved Learning of Academic Writing

451

• The modeling language must be usable by imple-

menting a ﬁrst symbol library (e.g. in Microsoft

Visio). Of course, handwritten sketches can still

be created beyond this.

• The language must be made known so that an ap-

plication is possible. This includes the presenta-

tion in lectures as well as the presentation in e.g.

a handbook.

• Typical conﬁgurations of functional elements and

possibly also presentation elements should be des-

cribed, which can be used as models for academic

text work in teaching. In this case, the modeling

language and the formulated patterns also serve

as learning materials for general requirements of

academic texts. (Figure 5).

EXAMPL HYPOTH ANALYS EVIDCOMP

Description

of an

example

stating the

problem

Formulating

a hypothesis

with its

different

aspects

Modeling

the situation

and/or the

process

Comparing

the relevant

parts with

known

approaches

Making

explicit the

support for

the further

stated

hypothesis

Figure 5: Argumentation Pattern as a Text Model.

• Processes are to be deﬁned which integrate the

development of the different design perspectives.

Here you will also ﬁnd references in the literature

describing, for example, the integration of outline

and content sketches in text planning. The use of

patterns or the connection with support processes

when creating ﬁnal theses must also be embedded

here.

A further challenge can be seen in the need for an in-

creased technical support for students or academics.

If it were possible not only to create text models in

the four perspectives deﬁned in this paper, but also

to integrate the text itself, the creation of a text could

be seamlessly designed from the draft to the writing

process. Here one could write in different parts of the

work, adapt the design and thus achieve an integrative

editing with an always good overview for the author.

Existing text creation tools allow to create a ﬁrst

outline of the text and then add the text. Thus, Micro-

soft Word also offers a so-called Outline View. Other

tools go one step further and not only integrate the

text with the separate outline representation, but also

integrate additional information for the writing pro-

cess into the document. For example, the tool Scrive-

ner (www.literatureandlatte.com/scrivener) contains

a research directory for collecting this information.

Scrivener especially offers the management of meta-

information in a binder, which refers to the text type

Novel (with characterization of ﬁgures etc.).

Figure 6 shows an initial draft of a scrivener-like

user interface that can integrate the perspectives of

a text model. While the well-known outline on the

left-hand side of the screen arranges the text’s content

components hierarchically, the functional and presen-

tation view can be found in the middle columns. Only

the section selected in the outline is displayed. From

the presentation elements, you can ﬁnd a direct assig-

nment to the actual text, which is located in the rig-

htmost column. The individual columns can also be

hidden by the user.

Finally, there are other use cases in the utiliza-

tion of text models and corresponding software tools

conceivable:

• In scientiﬁc practice, contributions are often writ-

ten in groups. This concerns the distribution and

coordination of the respective text contributions.

ANALYS

COMP

Figure 6: GUI prototype of a writing tool integrating text structure.

CSEDU 2018 - 10th International Conference on Computer Supported Education

452

It is also conceivable to use formal notation ele-

ments that deﬁne the expectations on both sides of

the respective text parts, similar to interface clas-

ses in software systems.

• The processes for the creation of texts with text

models have already been addressed and their

relation to the support processes. However, it

would also be conceivable to have guided lear-

ning processes that can be provided by the lecturer

with different speciﬁcations at all levels (students

could not only formulate texts but also analyse

them with regard to the functions they contain).

• Text models could not only support the creation of

texts, but also the reengineering of existing texts.

The aim of such an analysis can be the impro-

vement of texts in their creation process, the un-

derstanding of extensive and complex texts and

thus also the learning of writing techniques using

concrete examples. A software-technical support

of the analysis process is conceivable similar to

systems for qualitative data analysis (e.g. Atlas/ti,

www.atlasti.com). The possibility of greater auto-

mation when analysing texts should be investiga-

ted.

REFERENCES

Beaugrande, R. B. and Dressler, W. (1981). Introduction to

Text Linguistics. Routledge, London ; New York.

Bell, P. (1997). Using Argument Representations to Make

Thinking Visible for Individuals and Groups. In Pro-

ceedings of the 2nd International Conference on Com-

puter Support for Collaborative Learning, CSCL ’97,

pages 10–19, Toronto, Ontario, Canada. International

Society of the Learning Sciences.

Booth, W. C., Colomb, G. G., and Williams, J. M. (2003).

The craft of research. Univ. of Chicago Press, Chi-

cago, Ill. [u.a.].

Buzan, T. (2006). The Ultimate Book of Mind Maps. Harper

Collins Publishers, London.

Byrne, J. R. and Tangney, B. (2010). CAWriter: a

CSCW/CSCL tool to support research students’ aca-

demic writing. In Proceedings of the 24th BCS Inte-

raction Specialist Group Conference, pages 458–462,

Swinton, UK, UK. British Computer Society.

Cofﬁn, C., Curry, M. J., Goodman, S., and Hewings, A.

(2002). Teaching Academic Writing: A Toolkit for

Higher Education. Routledge, London.

Dijk, T. A. v. (1980). Macrostructures: an interdisciplinary

study of global structures in discourse, interaction,

and cognition. L. Erlbaum Associates, Hillsdale, N.J.

Evans, E. J. (2003). Domain-Driven Design: Tackling Com-

plexity in the Heart of Software. Addison Wesley, Bos-

ton.

Faustmann, G. (2011). Project-based Collaborative Lear-

ning with Arguments - Flexible Integration of Digital

Resources. volume 2, pages 359–364, Noordwijker-

hout, The Netherlands. SciTePress.

Hausdorf, C. (2005). Integratives, ganzheitliches Modell

und Werkzeug zur wissenschaftlichen Textpro-

duktion. PhD thesis, Dissertation. Erlangen-

urnberg http://www. opus. ub. uni-erlangen.

de/opus/volltexte/2005/125/.

Heard, S. B. (2016). The Scientist’s Guide to Writing: How

to Write More Easily and Effectively throughout Your

Scientiﬁc Career. Princeton Univers. Press, Princeton

Oxford.

Kirsh, D. (2010). Thinking with external representations.

AI & SOCIETY, 25(4):441–454.

Neuwirth, C. M. and Kaufer, D. S. (1989). The Role of

External Representation in the Writing Process: Im-

plications for the Design of Hypertext-based Writing

Tools. In Proceedings of the Second Annual ACM

Conference on Hypertext, HYPERTEXT ’89, pages

319–341, New York, NY, USA. ACM.

Newell, A. (1979). Reasoning, problem solving and deci-

sion processes: The problem space as a fundamental

category.

Shibata, H. and Hori, K. (2002). A framework to support

writing as design using multiple representations. In

Proc. 5th Asia Paciﬁc Conference on Computer Hu-

man Interaction. Citeseer.

Streitz, N., Haake, J., Hannemann, J., Lemke, A., Schu-

ler, W., Schtt, H., and Thring, M. (1992). SEPIA:

A Cooperative Hypermedia Authoring Environment.

In Proceedings of the ACM Conference on Hypertext,

ECHT ’92, pages 11–22, New York, NY, USA. ACM.

Toulmin, S. E. (2003). The Uses of Argument. Cambridge

University Press, Cambridge, U.K. ; New York, upda-

ted edition.

Improved Learning of Academic Writing

453