Print Email Download

Paid Writing Services

More Free Content

Get Your Own Essay

Order Now

Instant Price

Search for an Essay


Messages Segmentation of Asynchronous Text-Based Discussion in Social Network Analysis

I. Introduction

Online asynchronous discussion is now regularly employed in e-learning, and often archives of rich but unexploited information [1]. From the perspective of social network analysis, communication structure may be put through an objective analysis process to infer the outcome of the communication. In a teaching and learning context, the communication structure affords the revelation of the learning process, the learners' characteristics, and the learning outcome.

Existing social network analysis (SNA) is used independently in e-learning as a useful tool for studying relations. It is a collection of graph analysis methods that instructors and teachers used to analyze network interactions in their classes. Furthermore, SNA provides information regarding the communication structure of the group, level of participation, identifying who are central actors and other structural characteristics of the discussion forum [2].

However, in online discussion, using message as a basis for evaluating the participation and interaction among students is common. The instructors and teachers implemented SNA based on message, such as the number of message that posted by participants, but this is an unfair measurement for the participant who posts a message that contains several sentences.

The aim of this paper is to employ the application of message segmentation to support social network analysis for measuring the level of participation of online discussions in a few teaching and learning situations. The importance of development of segmentation procedure and its impact on the communication structures will be discussed.

II. Background

Social network analysis (SNA) is the study of the structure of social interactions [3]. It is based on the tenet that SNA focused on the study of structural information contained interactions between entities. Such information can therefore be inferred objectively and consistently from communication structure.

Studying the SNA of online discussion forums can serve a number of purposes. A common study purpose found in the literature is to understand various online teaching and learning issues through studying the online discussion structure. A component of such study is to identify students' perceived interaction in online discussion. For Example, [4] investigated the pattern of interaction in networked learning and computer supported collaborative learning; and [5] studied the students' interaction and meaning construction in computer bulletin board discussion. Another purpose of social network analysis is to assist in online learning discussion where the quantity of interaction is considered salient [6].

Social network analysis has also been applied in monitoring the learning progress of students. For example, [7] applied centrality and density to identify who are central actors in the network and the level of engagement among participants, and from it inferred the learning progress of the cohort.

Online discussions, using asynchronous message, is often used within an online class in e-learning. It has been shown that, within a web-based class, active online participation is a key factor in the success of student learning [8]. Furthermore, many instructors count the number of messages written by a student as a proxy for that student's participation [9].

However, assessing level of participation based on number of message could not help provide a clearer picture of what is happening in the online discussion. Moreover, it is not fair for participants who send or post a message that consist of many sentences, but recent advances in message segmentation have demonstrated the promise of this technique to support SNA based on number of sentences.

III. Research Design

We conducted centrality indicator to find the central participants within the network. With degree the network activity of individual members can be indicated. This can be done by calculating the in-degree and out-degree centrality measures. Secondly we conducted a density indicator to describe the overall linkage between the participants. Finally, both of indicators - degree centrality and density will be applied and compared in two type of unit interaction; message unit and sentence unit.

A. Message Segmentation

In order to automatically build the valid sentence of a message text, development of segmentation procedures is needed. We adopted Strijbos's segmentation procedure in [10] to create the message segmentation tool.

Message segmentation component allows the users to segment the text into sentences. Each message is first segmented in sentences by using full stop, question mark or exclamation mark that the author of the message has written. Each sentence that is followed by a 'full stop' constitutes a segment, regardless whether a 'finite form' or 'verb' is missing. Each compound sentence is split in segments using punctuation signs and symbols or signs that are used for punctuation purposes; comma, semicolon, colon, brackets, the word 'and', dash, and (...) or ...[10]. Non-English text should be excluded.

In this paper, we presented message segmentation tool in figure 1 for marking up the rhetorical structure of message. Message segmentation tool can automatically detect sentence boundaries. Clicking on the message segmentation button at the below of the layout automatically break each message into sentences.

From a message above consist of fifteen sentences. Number of these sentences will be applied in social network analysis tool as a basis of analysis unit to determine the level of student's participation.

B. Experiments

In this research experiment we will study two scenarios of social network analysis; (a) based on message and (b) based on sentence which is represented by a message segmentation experiment. A rule of message segmentation technique will be applied to one of this scenario and will be compared to other and the performance will be studied.

Experiment 1 stems from the requirement of identifying number of messages of online discussion to assess the level of students' participation. Usually, the researchers and the teachers using this way to analyze level of participation even though the message consist of only one word or one sentence.

Experiment 2 selected sentence as a single unit of interaction. The number of sentence breakdown from the message which is segmented during message segmentation process will be taken as an input to SNA. In these second experiments, message segmentation tool will be used to divide a message automatically into sentences.

The network analysis software-NetMiner [11] was used to conduct the network analysis and represent these online interactions experiments in visual object. Visual representation is important to understand the network data and convey the result of analysis.

C. Procedure

An online discussion forum from one subject SCK3433-02 2008/2009: Management of Organization Information Systems held on Moodle as a learning management system (LMS) in e-learning was examined for one thread. This thread was chosen because it has highest replies than other threads. The student population of 12 is well acclimatized with online discussion. A total of 38 messages have been posted and would be split into 116 sentences. The discussion was started by student A. The number of message and the number of sentence would be applied in SNA. Hence, represent this experiment in visual object to gain a better understanding of a student's involvement within online discussion.

IV. Result and Discussion

The following table shows the performance of SNA for both scenarios. Table 1 and table 2 shows the resulting adjacency matrix of interactions in the online discussion space is shown that showed which participants responded and posts to each other and how often they did so based on message and sentence respectively. The initial posts were not considered interaction and were not counted.

This is cumulative data: for example, in adjacency matrix based on message, it shows that the student E responded to student D one time during discussion and in adjacency matrix based on sentence, it shows that student E responded to student D also one time. It means that message's student E consist of one sentence or one word. Let's compare to adjacency matrix based on message that student K responded to student A one time, on the other hand, in adjacency matrix based on sentence, it shows that student K responded to student A five times. It means that message's student K consist of five sentences. Usually in SNA one sentence have same rate with five sentences.

Furthermore, centrality measures are being conducted to find central actors in a network to evaluate the student's participation. This can be done by calculating the in-degree and out-degree measure (table 3).

In this study in-degree provides information about the amount of messages or amount of sentences for student A when student A in a group receives messages or sentences from others in communicative situations (AßB). Out-degree gives an indication of the amount of messages or amount of sentences for student A when student A in a group sends messages or sentences toward others in communicative situations (AàB). Student A is high in-degree for both experiments indicates that his/her receives more information or comments from others and this student has more prestige in the network. Student H, high in out-degree based on number of message while student I and student L, high in out-degree based on number of sentence, indicates that his/her more active in providing information to others in discussion or providing comments on the opinions of others based on each perspective. Out-degree also describes the extent to which students learn actively. Student H post eight messages that consist of sixteen sentences (mean=2), on the other hand, student I and student L post four and five messages respectively that consist of nineteen sentences (mean=4.75 and 3.80).

Student A, student I and student L were the actors who ranked higher in both in-degree and out-degree considered so far. These students were prolific, consulted very often and mediated the flow information. However, the student that had lower rates in both dimension can be classified as lurkers or isolates.

In table 4 and table 5 reports on in-degree and out-degree based on message and sentence are shown respectively. We conducted density calculations to get an indication of the overall linkage of students in the network. This gives an indication of the level of engagement in the network. Density calculations indicate how active the students are involved in the discussion and show how dense is the participation within it.

TABLE IV. Report on In-Degree and Out-Degree based on message

Measures

Value

In-Degree

Out-Degree

Sum

38

38

Mean

3.167

3.167

Std. Dev

4.394

2.154

Min

0

1

Max

16

8

# of isolate

0

Network density

0.288

TABLE V. Report on In-Degree and Out-Degree based on Sentence

Measures

Value

In-Degree

Out-Degree

Sum

116

116

Mean

9.667

9.667

Std. Dev

15.913

6.018

Min

0

3

Max

60

19

# of isolate

0

Network density

0.879

In this case of sending and receiving the messages/sentences that were exchange through online discussion had a density of 28.8% within 38 messages and 87.9% within 116 sentences. It means that network density in this online discussion have good participation, however only a few posting but each posting contains elaborate idea or some comments and suggestions from each other. based on message and based on sentence respectively. It is clear from the graph based on message that one high extreme is measured for 1 student (A); and based on sentence, four high extremes are measured for 4 students (I, L, A and H), and they have an out-degree centrality of 1.727, 1.727, 1.545 and 1.455 respectively. They are the most powerful actors of the network and they are positioned toward the center of the out-degree centrality circle. They actively participate and provide information and comments on the opinions of others. They also have friendly relations with many students and have important roles in delivering information to their community. The main idea here is in based on message only student A high in out-degree, on the other hand, in based on sentence, there are four students high in out-degree and student A is in fourth position in out-degree centrality. It denotes that student A was posting many messages but the messages probably consist of one word or one sentence.

Table 6 shows the excerpts from the online discussion. From this table it is clear that one message can provide one sentence even one word, on the other hand, other provide several sentences. For example, student A presented one of his messages that consist of one sentence to emphasize his statements. However, other students engaged in more creative participation to deliver their ideas. For example, student I post or reply one of his messages that consist of five sentences to elaborate his views about the topic of the discussion. Hence, assessing students' interaction in online discussion based on sentence can give more objective way in communication structure.

TABLE VI. Excerpts from online discussion

Introduction, week 3

Message post or reply by students

Student

Name

Content

NOM

NOS

A

All right...

1

1

I

First of all I agree with the answer. For me, customer service is one of the main roles in this company. As the customer service employee, they have to be more efficient and committed to their job. But sometimes, there are errors made by them that cannot be avoided. According to wikipedia "Customer service is a series of activities designed to enhance the level of customer satisfaction - that is, the feeling that a product or service has met the customer.

1

5

L

Actually, the IS is a very common of meaning...So, refer to the functional of IS. I think it is about of theory for predicting the distribution of decision making between the corporate and business-unit levels of management for a subset of information systems (IS) resources referred to as systems development......

1

3

· NOM = Number of Message

· NOS = Number of Sentence

Table 7 shows different result produce by the two (message and sentence) unit interaction in term of SNA indicators. Mean of degree by sentence is more realistic compare to mean of degree by message. This can be useful to understand how much of the dialog among the students based on these two types of interaction.

TABLE VII. Comparative result: based on # of Messages and based on # of sentences

Measures

Based on # of

Message

Based on # of

Sentence

Total

38

116

Mean

3.167

9.667

Student highest in degree

· In-Degree

· Out-Degree

A (16)

H (8)

A (60)

I and L (19)

Student A à sixteen messages = sixty sentences (mean=3.75)

Student H à eigth messages = sixteen sentences (mean=2)

Student I à four messages = nineteen sentences (mean=4.75)

Student Là five messages = nineteen sentences (mean=3.80)

Student highest in centrality

· In-Degree Centrality

· Out-Degree Centrality

A (1.455)

H (0.727)

A (5.455)

I & L (1.727)

Network density

0.288

0.879

It also shows who is the highest in degree and centrality. The result might be different between using the number of message and using the number of sentence. For example, student H is the highest in out-degree based on message while student I and student L, are the highest in out-degree based on number of sentence. It means that message by student A only consist of few sentences compare to student I and student L which contribute longer messages (more than one sentence).

V. Conclusion

This paper proposed sentences as the unit interaction instead of message to assess the level of participation among students. The purpose of employing sentence for counting number of interaction is to emphasize that student can post a message regardless of the length. It is not fair for a participant that post or reply a message that consist of many sentences that represent many ideas but only being considered as one contribution of participation. Participants that post or reply many sentences and post or reply few sentences should not be treated equivalent level of participation. Those who participant that post long message that represent many ideas should be recognize to contribute more in the participation.

In addition, we also describe the message segmentation that can be used to support SNA in online discussion. With the message segmentation tools, a message can be broken down into sentences, which able to preserve the strengths of social network analysis in communication structure.

References

[1] A.K.F. Lui, S.C. Li, and S.O. Choy, An evaluation of automatic text categorization in online discussion analysis, Seventh IEEE International Conference on Advanced Learning Technologies, Niigata, Japan, 2007, 205-209.

[2] Erlin, Y. Norazah, and A.R. Azizah, Students' interaction in online asynchronous discussion forum: A social network analysis, International Conference on Education Technology and Computer, Singapore, 2009, 25-29.

[3] S. Wasserman, and K. Faust, Social Network Analysis, Cambridge University Press, 1994.

[4] M. de Laat, V. Lally, L. Lipponen, and R.J. Simon, Investigating patterns of interaction in networked learning and computer-supported collaborative learning: A role for Social Network Analysis, Journal of Computer Supported Collaborative Learning, 2(1), 2007, 87-103.

[5] J.B. Pena-Shaff, and C. Nicholls, Analyzing student interactions and meaning construction in computer bulletin board discussions, Computers & Education, 42(3), 2004, 243-265.

[6] H.J. Suh, and S.W. Lee, Collaborative learning agent for promoting group interaction, ETRI Journal, 28(4), 2006, 461-474.

[7] P.A.Willging, Using social network analysis technique to examine online interaction. US-China Education Review, 2(9), 2005, 46-56.

[8] S.R. Hiltz, and M. Turoff, M, What makes learning networks effective?, Communications Of The ACM, 45(2), 2000, 56-59.

[9] R. Vinaja, and M. Raisinghani, Analysis of strategies used in teaching an online course in a predominantly hispanic university, JCSC, 2001, 70-79.

[10] J.W. Strijbos, R. Martens, J. Prins, and W. Jochems, Content analysis: What are they talking about?, Computer & Education, 46(1), 2006, 29-48.

[11] http://www.netminer.com/