A GENERIC MODELING OF INDICATOR WITH UTL

The Collaborative Action Function Example

Sébastien Iksal, Christophe Choquet and Diem Pham Thi Ngoc

Computer Science Laboratory, University of Maine, Le Mans, France

Keywords: TEL Engineering, Activity Tracks Modeling, Modeling Language, Capitalisation, Reusing.

Abstract: Capitalising the know-how of educational data analysis techniques in order to share and re-use them is one

of the major stakes today in the specific Technology Enhanced Learning (TEL) field. We have proposed a

Usage Tracking Language (UTL) to formalize data analysis techniques with a goal of capitalisation and

reusability. In this paper, we present an example of use on the collaborative action function which is an

indicator proposed in the ICALTS Project. After a short presentation of UTL and its conceptual model, we

introduce the declarative Data Combination Language for UTL called DCL4UTL and we detail the

Collaborative Action (CA) indicator modeling.

1 INTRODUCTION

Educational Data Mining (EDM) techniques are

often complex. Their implementation and use for a

specific Technology Enhanced Learning (TEL)

system is time spending. Thus, as many authors

claims (see for example (Mostow, 2004)), one of the

major bottlenecks in this field is to find solutions for

easily sharing and reusing such techniques. With this

in mind, the EU research community has driven five

significant research actions for exploring the

European know-how in this field (“ICALTS, 2004),

(“Interaction Analysis”, 2005), (“TRAILS”, 2004)

and for proposing solutions for the capitalisation and

the reuse of EDM and data analysis techniques

(“CAViCoLA”, 2006), (“DPULS”, 2005).

With the help of these projects results, we have

proposed the Usage Tracking Language (UTL) for

formalising data mining and data analysis techniques

and for capitalising them as Design and Analysis

Patterns (Choquet and Iksal, 2007). Recently, we

have proposed a new version of this language where

the automatic methods for collecting data from a

TEL system and for elaborating significant

indicators (in the meaning of (Harrer, Martinez-

Mones and Dimitracopoulou, 2009)) for the users

(researchers, TEL designers, teachers, students)

could be expressed in a declarative and generic way,

and could be interpreted for mining tracks from

different TEL systems (Pham Thi Ngoc, Iksal and

Choquet, 2009).

This paper will present how an indicator could be

modelled and performed with UTL. Our aim is to

demonstrate that the modeling of the indicator and

its calculation method is (1) generic and reusable,

and (2) understandable by humans who want to use

it in a TEL system and computational, that is to say,

understandable by computers. We have chosen a

well-known indicator of collaboration in a

synchronous Computer-Supported Collaborative

Learning (CSCL) system to exemplify the use of

UTL: the Collaborative Action function Indicator

(CA) which was initially defined and implemented

in the MODELLINGSPACE platform (Avouris,

Dimitracopoulou, Komis and Margaritis, 2003).

The calculation of this indicator is based on two

others, the Agent and the Interaction indicators. As

synthesised below, the Agent indicator computes in

a CSCL system the number of users (learners,

teachers) who have used at least one collaboration or

interaction tool during a collaborative session. The

Interaction indicator expresses the amount of

interactions that have implemented through a

communication channel (e.g. a chat or any

collaborative workplace) during a given period.

Finally, the Collaborative Action Function indicator

combines these indicators for providing a

quantitative evaluation of the collaboration inside a

group of users during a given period.

Consider a collaboration session interval [t

–t

This time interval is quantified using a parameter n:

= t

+ i*d, where d = (t

- t

)/n. Then CA is expres-

114

Iksal S., Choquet C. and Pham Thi Ngoc D. (2010).

A GENERIC MODELING OF INDICATOR WITH UTL - The Collaborative Action Function Example.

In Proceedings of the 2nd International Conference on Computer Supported Education, pages 114-119

 SciTePress

sed as follow.





maxk

iii )tI(k,*)tA(k, )CA(t

(1)

The next section of this paper presents the

conceptual model of UTL. The third section

describes the modeling of the Collaborative Action

Function indicator with UTL. We concludes this

paper by dealing with the benefits and limitations of

the use of UTL regarding the capitalization, the

sharing and the reuse of indicators.

2 THE USAGE TRACKING

LANGUAGE: UTL

As we said before, UTL was proposed in order to

formalise data mining and data analysis techniques

in a declarative way for capitalising and re-using

them. We first present the DGU Model on which

UTL is based for its descriptions, next, we introduce

the conceptual model of our language and we finish

with the combination language which is used to

describe the data transformation and calculation

methods.

2.1 Three Viewpoints on Tracks:

The DGU Model

In our tracks' reengineering framework, we stress the

need for track modeling before the learning session,

and to consider tracks as a pedagogical object, like

any other learning object, such as resources or

scenarios for instance. If this is frequently the case

in existing systems, when the tracking purpose is to

provide learners and/or the system with useful

information, it is more unusual for supporting the

tutor, and it's rare for providing feedback to

designers.

With this in mind, we could say that, as far as it

could be possible for him, the designer who is

engaged in the engineering process of a TEL system,

should model the tracking needs of the learning

session. Then, the tracking needs should be

instantiated on the educational environment in order

to model the effective tracking means. Finally, one

should also model the expected uses of these tracks,

in terms of building the descriptive scenario of the

learning session for analysing the usage of the TEL

system. This is the way we have defined the three

facets for the tracks modeling (see Figure 1):

– the Defining (D) facet which models the tracks

needs;

– the Getting (G) facet which models the tracks

means;

– the Using (U) facet, which models the tracks

uses purpose.

Figure 1: The DGU model of a Track.

2.2 The Conceptual Model

UTL Conceptual Model falls into three parts (cf.

Figure 2), but we have restricted the focus of this

paper on its re-usable part. This part of the

conceptual model is called UTL/P (for

UTL/Pattern), and corresponds to the

capitalisable/reusable part of the language, details all

transformations and data needed to build an

indicator. We have identified two main data types

for tracks: the derived-datum and the primary-

datum.

derived-datum

primary-datum

intermediate-datum

indicator

additional-datum

tracking-purpose

track

use

is relevant of

raw-datum

is modeled by

content-datum

datum

defining

getting

using

UTL/P

UTL/S

UTL/T

traceable-concept

traceable

Figure 2: UTL Conceptual Model.

The primary data are not calculated or elaborated

with the help of other data. They consist of raw-

datum, content-datum and additional-datum. The

raw-datum could be recorded before, during or after

the learning session by the learning environment.

The content-datum concerns to the outcomes

provided by the learning session's actors. The

additional-datum is linked to the learning situation

and could be involved in the usage analysis.

The derived data are calculated or inferred from

primary data and/or other derived data. The indicator

is a derived-datum which has a pedagogical

significance and is always relevant to a pedagogical

A GENERIC MODELING OF INDICATOR WITH UTL - The Collaborative Action Function Example

115

context. It is always defined for, at least, one

exploitation purpose (for instance, in an instructional

design context, we have proposed to link each

indicator to at least one concept of the learning

scenario). A derived-datum which has to be

calculated but which has no pedagogical

significance is qualified as an intermediate-datum.

All the information models of these data types have

been published in (Choquet and Iksal, 2007).

2.3 The Data Combination Language

In order to build derived data and indicators we have

proposed a data combination language extension

called DCL4UTL (Pham Thi Ngoc, Iksal and

Choquet, 2009). The main aim of this language is to

be used by the analyst as a query language to

describe the way to combine data for producing a

new one. This language is designed to be declarative

and to be processed automatically by the system.

Another research goal concerns the capitalisation of

these combination methods for the re-usability of

them. Our proposal is based on classical languages,

integrated in UTL and works on the Using section of

UTL data. All Using sections have a format field

and a data field. The format describes how the value

has to be represented, and the data field concerns the

result of the combination. The grammar of the

language is too long to be published here, thus we

focus on a UTL example of use.

3 AN EXAMPLE: THE

"COLLABORATIVE ACTION

FUNCTION" INDICATOR

To elaborate compatible raw data with the

calculation of the CA indicator, we assume the TEL

system is able to provide logs with the following

information:

{logfile}:{time stamp, actionCode,

user, resourceID}

Where:

– actionCode corresponds to the interaction

performed,

– user corresponds to the agent involved,

– resourceID corresponds to the interaction

channel used.

We compute CA(ti) by parsing that kind of log

files after or during the learning session.

3.1 Some Explanations on the

DCL4UTL Syntax

DCL4UTL is based on classical operators:

– Logical operators : and, or, not, …

– Comparison operators : <=, >=, =, …

It is possible to use some pre-defined functions, in

this example, we need the following:

– 'first' and 'end' to get the first and the last

elements of a set of data.

– 'count' and 'countDis' to get the number of

elements and the number of distinct elements of

a set of data.

– 'sort' to organise a set of data according to a

criterion.

– 'filter' to extract a subset of data according to a

criterion.

In order to organise the combination, we have the

following keywords:

– 'cal' to declare the combination

– 'as' to declare the specific organisation of data

(filter, sort, …)

– 'for each' is a control for a loop.

– 'where' allows aliases for data.

3.2 Modeling of the Collaborative

Action Function Indicator

This section presents the Defining (Table 1) and the

Using (Table 2) parts of the UTL modeling. They

correspond to descriptions that are independent from

the calculus which is detailed in the next section.

Table 1: Defining Section of CA Indicator.

Defining

Title

Collaborative Action Function

(CA)

Cardinality

Description

Indicates the degree of

collaborative action […].

Table 2: Using Section of CA Indicator.

Using

Data

Not used in modeling

Format

float

</ca>

</indicator>

3.3 Calculation Method of the

Indicator

In this section, we will not present all data modeling,

CSEDU 2010 - 2nd International Conference on Computer Supported Education

116

but the format for primary data and for derived data,

we will give the format and the getting formula.

1. Firstly, we create the RD-AUR raw datum (in

the UTL meaning) to model the log file's

content.

<actionCode> string </actionCode>

<user> string </user>

<resourceID> string </resourceID>

</rawDatum>

2. Secondly, we describe the additional data (in the

UTL meaning) corresponding to the number of

intervals and the list of interaction channels:

a) The AD-Parameter additional datum stands

for the parameter n, which is the number of

intervals.

<n> integer </n>

</additionalDatum>

b) The AD-Channel additional datum stands for

the interaction channels' set.

<channel id="integer"> string

</channel>

</additionalDatum>

3. To compute CA(t

), I(k,t

) and A(k,t

), we need to

define and calculate (in the UTL meaning) the

following intermediate data.

a) The ID-BET intermediate datum stands for

the beginning and the ending time of the

session.

</intermediateDatum>

The calculation expression is:

cal ID.t0 = first(R.timeStamp);

ID.tm = end(R.timeStamp)

as sort(R.timeStamp)

where ID = ID.BET.using.data ┐

.intermediateDatum,

R = RD-AUR.using.data.rawDatum

The ID-D intermediate datum stands for d=(t

- t

)/n.

<d> float </d>

</intermediateDatum>

The calculation expression is:

cal ID1.d = (ID2.tm - ID2.t0) / AD.n

where ID1 = ID-D.using.data ┐

.intermediateDatum,

ID2 = ID-BET.using.data ┐

.intermediateDatum,

AD = AD-Parameter.using.data ┐

.additionalDatum

4. Then, we compute two indicators: I-Interactions

and I-Agents.

a) The I-Interactions indicator stands for the

amount of interactions that have been

implemented via channel k during (t

i-1

-ti ].

<interactions ti="time"

channel="integer">

integer

</interactions>

</indicator>

The calculation expression for all interaction

channels is as follows:

cal for each $k in AD.channel[id] do

for each $value in [1;A.n-1] do

$ti1 = ID1.t0 + $value * ID2.d;

$ti0 = ID1.t0 + ($value-1) *

ID2.d;

I.interactions[ti=$ti1]

[channel=$k] =

count(R.actionCode)

as filter(R.timeStamp <= $ti1 and

R.timeStamp > $ti0 and

R.resourceID == $k)

where I = I-Interactions.using.data ┐

.indicator,

A = AD-Parameter.using.data ┐

.additionalDatum,

AD = AD-Channel.using.data ┐

.additionalDatum,

ID1 = ID-BET.using.data ┐

.intermediateDatum, ID2 =

ID-D.using.data ┐

.intermediateDatum,

R = RD-AUR.using.data.rawDatum

b) The I-Agents indicator stands for the number

of agents that have posted at least one

message through channel k during (t

i-1

-ti]

interval.

<nbagents ti="time"

channel="integer">

integer

</nbagents>

</indicator>

The calculation expression for all interaction

channels is as follows:

cal for each $k in AD.channel[id] do

A GENERIC MODELING OF INDICATOR WITH UTL - The Collaborative Action Function Example

117

Table 3: Getting section of the CA Indicator.

Getting

Title

Calculation of the Collaborative Action Function

Component

Primary-datum

AD-Channel

Component

Derived-datum

ID-D

Component

Derived-datum

ID-BET

Component

Derived-datum

I-Agents

Component

Derived-datum

I-Interactions

Method

Type

Automatic

Tool

Title

Declarative combination function based on agents and interactions data

Description

Text

CA(t

)=Sum[k=1;kmax](Agents(k,t

)*Interactio

ns(k,t

))

DCL4UTL-formula

Cf. formula located after this table

for each $value in [1;A.n-1] do

$ti1 = ID1.t0 + $value * ID2.d;

$ti0 = ID1.t0 + ($value-1) *

ID2.d;

I.nbagents[ti=$ti1][channel=$k]

= countDis(R.user)

as filter (R.timeStamp <= $ti1

and R.timeStamp > $ti0 and

R.resourceID==$k)

where I = I-Interactions.using.data ┐

.indicator,

A = AD-Parameter.using.data ┐

.additionalDatum,

AD = AD-Channel.using.data ┐

.additionalDatum,

ID1 = ID-BET.using.data ┐

.intermediateDatum,

ID2 = ID-D.using.data ┐

.intermediateDatum,

R = RD-AUR.using.data.rawDatum

5. Finally, we compute CA(t

) by defining indicator

I-CA. The Table 3 represents the UTL modeling

of the Getting part.

cal for each distinct $ti in

I1.interactions[ti] do

I.ca[ti=$ti] = 0;

for each $k in AD.channel[id] do

I.ca[ti=$ti] = I.ca[ti==$ti]

+ (I1.interactions[ti==$ti]

[channel==$k]*I2.nbagents

[ti==$ti][channel==$k])

where I = I-CA.using.data.indicator,

I1 = I-Interactions.using.data ┐

.indicator,

I2 = I-Agents.using.data ┐

.indicator,

ID1 = ID-BET.using.data ┐

.intermediateDatum,

ID2 = ID-D.using.data ┐

.intermediateDatum,

AD = AD-Channel.using.data ┐

.additionalDatum

4 CONCLUSIONS

With this paper, we have shown how UTL and its

extension DCL4UTL could be coupled for

expressing and modeling indicators in a generic and

reusable way. Each datum required for modeling an

indicator and each calculation method could be

expressed in a generic way. Part of their description

targets the human (and more specifically the teacher,

as the designer or as the tutor), as getting.example or

getting.method.tool.description.text fields, for

facilitating the capitalisation and the sharing of an

indicator, expressed as a Pattern. Some other fields

targets the computer for allowing the reusability and

the interoperability, as using.format or

getting.method.description.DCL4UTL-formula.

The only TEL system specifics are the raw data

extraction methods which have to be coded

according to the specificity of the log files.

Even if the major part of UTL is understandable

by a teacher, it is not the case of DCL4UTL

extension we propose: using this language requires

competencies of a data analyst and/or a software

engineer. This is one the limitation of our approach:

actually, a teacher could not be independent during

the indicators modeling task, he should be helped by

specialists. Nevertheless, we think that this

limitation could be bypassed at middle-term: more

indicators will be modelled and capitalised with

UTL, more they can be shared in the community of

practice and reused by end-users, e.g. teachers.

We have actually developed the interpreter of

DCL4UTL, and improved it on existing indicators

we found in the literature and/or we need for our

existing TEL systems (see for instance (Pham Thi

Ngoc et al., 2009)). We have planned also to

develop a user-friendly editor for facilitating the

CSEDU 2010 - 2nd International Conference on Computer Supported Education

118

modeling with this language. We have also proposed

a Web server architecture for storing and sharing in-

dicators and other data modelled with UTL.

UTL could be tailored for different uses. As an

example, we have defined add-ons to UTL

specifically for Learning Management Systems

(LMS) such as the Moodle platform where the

learning activities are designed by using an

Educational Modelling Language (Rawlings, van

Rosmalen, Koper, Rodriguez-Artacho and Lefrere,

2002). Basically, one add-on, the UTL-Track (see

Figure 2) allows to model raw-data produced by the

LMS. The other one, the UTL-Scenario allows to

model links between indicators and the pedagogical

scenario. In this way, an indicator could be put in the

context of a specific concept of a scenario (as a

learner role, or a specific resource) and/or limited to

a specific passage of the learning activity.

We actually work on a new add-on tailored for

adaptable TEL systems. In this kind of system, the

user could take decisions of adaptation which impact

on the configuration of the TEL system and on the

pedagogical scenario itself. The add-on will allow to

model the context of such a decision, the modeling

phase could be held during or after the learning

session. This functionality will be useful for

analysing the pertinence of an adaptation and its

impact, for reengineering purposes.

ACKNOWLEDGEMENTS

We would like to thank all of the contributors of

CAViCoLA, DPULS, IA, ICALTS and TRAILS

European projects for their results reused here as

basis or examples.

REFERENCES

Avouris, N., Dimitracopoulou, A., Komis, V., Margaritis,

M., 2003. Participatory Analysis of Synchronous

Collaborative Problem Solving using the OCAF

Methodology and Tools. In Proceedings of CSCL2003

Community events. p. 232-234.

CAViCoLA, 2006. Computer-based Analysis and

Visualization of Collaborative Learning Activities,

European Network of Excellence Kaleidoscope. At:

http://www.noe-kaleidoscope.org (01/10/09).

Choquet, C., Iksal, S., 2007. Modeling Tracks for the

Model Driven Reengineering of a TEL System. In

Journal of Interactive Learning Research (JILR),

Special Issue Usage Analysis in Learning Systems:

Existing Approaches and Scientific Issues, Vol. 18

n°2. p. 161-184.

DPULS, 2005. Design Patterns for Recording and

Analysing Usage of Learning Systems, European

Network of Excellence Kaleidoscope. At:

http://www.noe-kaleidoscope.org (01/10/09).

Harrer, H., Martinez-Mones A., Dimitracopoulou A.,

2009. Users’ data: Collaborative and social Analysis,

In Technology-Enhanced Learning – Principles and

Products, Eds: Nicolas Balacheff, Sten Ludvigsen,

Ton de Jong, Ard Lazonder, Sally Barnes, Lydia

Montandon – Dordrecht : Springer Verlag, Chapter 11,

p. 175-193.

Interaction Analysis, 2005. Interaction Analysis –

Supporting participants in technology based learning

activities, European Network of Excellence

Kaleidoscope. At: http://www.noe-kaleidoscope.org

(01/10/09).

ICALTS, 2004. Interaction & Collaboration Analysis'

supporting Teachers & Students' Self-regulation,

European Network of Excellence Kaleidoscope. At

http://www.noe-kaleidoscope (01/10/09).

Mostow, J., 2004. Some useful design tactics for mining

ITS data, In Workshop on Analyzing Student-Tutor

Interaction Logs to Improve Educational Outcomes,

International Conference on Intelligent Tutoring

Systems (ITS'04). p. 20-28.

Pham Thi Ngoc, D., Iksal, S., Choquet, C., Klinger, E.,

2009. UTL-CL: A Declarative Calculation Language

Proposal for a Learning Tracks Analysis Process, In

The 9th IEEE International Conference on Advanced

Learning Technologies (ICALT' 2009).

Rawlings, A., van Rosmalen, P., Koper, R., Rodriguez-

Artacho, M., Lefrere, P., 2002. Survey of educational

modelling languages (emls). Technical report,

September 19, 2002.

TRAILS, 2004. Personalised and Collaborative Trails of

Digital and Non-Digital Learning Objects. European

Network of Excellence Kaleidoscope. At:

http://www.noe-kaleidoscope.org (01/10/09).

A GENERIC MODELING OF INDICATOR WITH UTL - The Collaborative Action Function Example

119