Improving browsing usability for mobile learners

Dimitrios P. Glaroudis Applied Informatics

University of Macedonia Thessaloniki, Greece

dglaros@uom.gr

Isabella Kotini Applied Informatics

University of Macedonia Thessaloniki, Greece

ikotin@uom.gr

Athanasios Manitsaris Applied Informatics

University of Macedonia Thessaloniki, Greece

manits@uom.gr

Abstract— Educational web sites can be used as supporting learning tools for students that wish to supplement their knowledge without restrictions of time and place. The objective of this work is to propose a methodology for improving the mobile learners’ web experience and provide personalized presentation of educational content. The proposed approach results in recommending relevant links to the users’ interests, by exploiting the logs of the educational portal and the semantics of the learning content.

Keywords- mobile browsing, content mining, ontology, personalization.

I. INTRODUCTION The evolution of Internet and multimedia services has

lead to interesting applications in education. Nevertheless, the acquisition of knowledge presupposes the existence of computers and internet access. Most times, however, owning or using a computer with a reliable Internet connection can be an important restriction for learners. Mobile communication technologies can overcome such restrictions, and prepare learners for more flexible training environments.

Because of economic realities, educators are compelled to consider new delivery strategies and approaches to developing lifelong learners. Recent discussions of higher education would seem to portray a view that the aims of higher education are consistent with the demands of working life. Development of personal skills by learners includes interaction of theoretical and practical knowledge, communication and cooperation skills, added by the ability to manage information and flexibility to work in different contexts or conditions. However, learners and Internet users generally, often cope with difficulties such as distraction, lost-in-space syndrome and cognitive overhead, when trying to acquire and access learning material. By knowing how learners gather web information and choose their navigation paths, recommendation systems can be modeled and partially

automate and predict learners’ steps, thus improving learning material discovery and simplifying management. Although there are different views on which features are relevant to Knowledge discovery by learners, the majority of researches [1, 2, 3] show that learners are highly goal-driven when they search information on educational sites. Thus, it is essential to improve browsing usability for learners with less effort and frustration, and, more important, by saving browsing time.

Enhanced management and browsing usability could be extremely useful for mobile web visitors, who browse content from cell phones or wireless PDA. Examining the way of use and the extensive popularity of mobile devices, it is easily understood why mobile learning can enhance the learner’s communication and cooperation skills, or increase their flexibility to perform in different working or learning environments. Mobile learning is defined in [4] as the learning process in which the user does not find itself in predetermined locality. Trifonova and Ronchetti [5] defined mobile learning as e-learning that can be transmitted via portable appliances, and more concretely via each appliance that has small size, is autonomous, accompanies the user each moment and can be used as a mean of transmitting or accepting educational material. Mobile learning can be spontaneous, personal, informal, contextual, portable, ubiquitous, and pervasive (so integrated with daily activities that are hardly noticed). Along with these attributes, mobile learning has much in common with other types of e-learning on desktop computers, but with the advantages of more immediate interaction and smaller, often wireless devices. Moreover, mobile devices cover, and predispose for further growth, the need of training community for social interaction and safeguarding of individuality [6, 7]. The exchange of data and the collaboration with other students is straightforward, therefore learners can be grouped and socialize more easily. Furthermore, mobile appliances assemble and deliver data in real time, offer information

262978-1-4244-5328-3/10/$26.00 ©2010 IEEE

about place and environment, increasing the sensitivity and the wish of students for experiments and real research activities. Besides, the individuality of users is safeguarded because they consider the accepted knowledge as personal. From the above, it is obvious that using mobile technology, knowledge can be acquired without time or territorial restrictions. Hence, mobile learning is gaining growing attention by researchers and scientists.

Besides their important benefits, mobile appliances have significant restraints when they are used for learning purposes. Sharples [7] formulated them as: limited memory and storage capacity, small screens, lack of continuous Internet connection, significantly browsing latency, and weakness to incorporate easily current learning services for desktop PC. These restraints have high impact on mobile users' browsing behavior. Studies have indicated that mobile Internet users suffer more severely from the problem of undesired outcomes than stationary Internet users do [8]. The limited screen size forces most mobile browsers to support a line-based navigation, with a few soft keys such as OK, Clear, and Next. In their design guidelines for small screen search engine interfaces, Jones et al. [9] clearly denote that screen size has a major impact on mobile users’ web experience. Specifically, success rates drop and the time to complete successful searches increases when mobile users navigate to the majority of websites. Additionally, they observed that users do not usually look at the entire content of search result pages. Furthermore, researches [8, 9, 10] for both desktop and mobile users show that page-to-page navigation is very costly when browsing in general. The authors of [3, 7, 8] advise web developers to reduce the amount of page-to-page navigation needed to view search results and to provide a quick way for users to navigate through conventional or small screen web pages.

Taking into account the above issues, this work aims to address the problem of how mobile learners can be efficiently supported by an educational platform in order to overcome browsing limitations. Working on this direction, a new methodology that aims to reduce mobile users’ effort when navigating to the educational portal and drive them to their desired educational material, while gaining knowledge without spatial and temporal constraints, is proposed.

The rest of this paper is organized as follows: In Section II an overview of the research efforts in the area of mobile browsing and web personalization is given. The main steps of the methodology and the particular methods and techniques are presented in Section III. The advantages and limitations of the proposed approach, along with conclusions are presented in Section IV.

II. RELATED WORK Lately, many research efforts concentrate on mobile

browsing usability. Present browsing methods for mobile devices can be classified into three main categories [11]: presentation optimization, semantic conversion, and scalable (zooming) methods.

Chen and Mohapatra proposed the method of scalable browser [11] as a new way to design personal digital

assistant (PDA) interfaces. However, the claim of authors that the proposed method is better over other prevailing methods, such as presentation optimization and semantic conversion, is not supported adequately with relevant experimental evidence.

Buyukkokten et al. [12] present important ideas for extracting semantics from the Web text yet greatly shortening the length of text. Each text page is intended to be broken into a number of text units that can be hidden, partially displayed, fully visible, or summarized.

Additional research has been conducted on dynamic text presentation for mobile devices using Rapid Serial Visual Presentation (RSVP) [13]. Moreover, researchers have proposed useful ideas on Web interface design, but have not performed sufficient experiments to compare different presentation methods in terms of their space efficiency.

Hierarchical menu structure has been used in user interface design based on spatial organization of information [14]. Additionally, content hierarchy or Hierarchical Atomic Navigation has been proposed as a new philosophy to improve Web navigation on small displays [15]. The idea is to divide an original page into zones and make the navigator page a reduced overview of the original page.

Henricksen et al. [16] address types of adaptation that can be applied to a Web browser in response to diverse context changes, including changes in the user context, input and output device capabilities, etc. Introducing an approach on the personalization-based optimization of Web interfaces for mobile devices, Hinz et al. [17] uses the ideas of structure adaptation and content adaptation to realize adaptive intelligent user interfaces. Additionally, in [18] and [19], a series of tools for desktop web browsing that create user profiles at server or client side and then provide feedback to the portals have been proposed.

Mobasher et al. [20, 21] proposed content characteristics and navigation data to be integrated into a Web mining framework and used by machine recommendations uniformly for more effective personalization. Thor and Rahm [22] followed this approach and developed an adaptive website recommendation system for use in e-commerce sites. Besides the fact that recommendations are critical to the success of large websites, there are different ways to define them and the quality of recommenders depends on many factors that remain unknown.

Finally, Eirinaki et al [23] worked on a different approach using the same basic logic. They proposed a system that uses log files and content semantics of a website in order to enhance personalization process. They introduce C-logs, an extended form of Web usage logs that encapsulates knowledge derived from the link semantics. C-logs are used as input to the Web usage mining process, resulting in a set of recommendations.

Besides the fact that content’s semantic conversion and personalization strategy is proposed in our approach, as happens also in some of the above mentioned works, a definition of the recommendation factors and a measurement of their quality, which is correlated with the user’s navigational patterns, is also defined. The proposed architecture makes use of both the usage and the educational

263

content in order to personalize the content’s presentation to mobile users. The innovative contribution of this work lies in identifying the mobile users who visit the learning management system (LMS) and their navigational patterns, while the LMS’s data is semantically enriched with ontology concepts. After tracking mobile users’ preferences from log records during past visits to the portal, an online system recommends a set of links to LMS web pages with similar conceptual content.

III. PROPOSED METHODOLOGY Educational site personalization to mobile users based on

their log files and conceptual analysis of educational content is proposed in our approach. If a system relies solely on usage-based records, then valuable information conceptually related to what is finally is worth to be recommended may be lost. However, after defining and modeling the semantics of the LMS content and simultaneously collecting mobile users’ navigation patterns, a knowledge system can then recommend personalized and desired knowledge to mobile learners. Therefore, the overall system (Figure 1) can analyze the information in web pages visited by the user, locate pages that have similar information and propose their links to the user. In the following section, the main phases of the proposed methodology are presented.

Figure 1. System architecture

A. Methodology’s main phases The proposed methodology is divided in four main

phases: Phase 1: Mobile user identification

Step 1: Data filtering Step 2: Session identification Step 3: User profile creation

Phase 2: Educational content mining Step 1: Content preprocessing Step 2: Text mining Phase 3: Knowledge system development

Step 1: Concepts definition Step 2: Concepts modeling Step 3: Evaluation process

Phase 4: Link suggestion

B. Methodology’s detailed presentation In this subsection the phases and subsequent steps of the

proposed methodology are presented along with particular methods and tools applied.

1) Mobile user definition Users’ navigational/usage data is logged in the web

server that supports the LMS and reveals the web pages that every user visits while browsing the portal. Although accessing the Web server log file (usage data) can be easy, the extraction of the actual browsing path in the form of structured information is a complicated and difficult process.

Initially, data filtering techniques have to be applied. The Web server log file stores all the requests and replies between servers and clients. However, a prime concern is to track only the so called useful data. Useful data are concerned with the actual web links of the pages that are being accessed, the IP addresses and the kind of clients’ web browsers, and the request times. Using a Java NetBeans environment that runs in the application server, the Web server log file is processed and data filtering is initially performed. The main concern is to use an open source tool with neutral architecture, meaning that it can be applied to portals and clients using different operating systems. After the initial data is filtered, the mobile user’s identification is established. In our approach, each user is defined by his IP address and the browser he uses. We consider only IP addresses and browsers that refer to mobile devices. This mode can significantly improve the speed and reliability of the developed system.

264

The second step involves the identification of mobile learners’ navigational patterns. This process is also established in the application server. Each user’s filtered data are analyzed and a session identification technique is used so as to reveal the user’s navigational pattern. User session is considered to be a sequence of requests coming from the same user and divided by a predefined period of time (timeout). Finally, each user’s data, serving as the user’s profile, are aggregated in order to form a knowledge base. These data are stored in the database server to be used in the following phases.

2) Educational content mining

After the user profile has been created, semantic

enrichment of the learning content is performed. However, prior to applying semantic conversion techniques, the preprocessing of the educational portal is essential in order for the content to be properly formatted. Java Net Beans environment combined with open source libraries runs in the application server to parse all the pages of the educational portal. Utilizing proper Java classes, only the educational content of the LMS is captured and copied in the database server. This step results in faster and easier content processing since the Web server hosting the LMS may host extra, irrelevant to the LMS data. Moreover, if further content will be added in the LMS, only the additional data will be processed.

In the second step of this phase, knowledge mining techniques are applied. In general, knowledge mining uses various recovery information methods (text mining, link mining, screen data selection or removal [24], [25]) without preserving the original web page structure. The choice of a mining method is directly related to how the developer wishes to acquire, analyze, and model the educational content. As there are many open source tools for ontology creations that work more easily and efficiently with simple text files, text mining for knowledge extraction has been chosen. Using proper open source Java classes, only the text information from each web page is extracted and the final information is formatted as a text string. Eventually, every page of the LMS is represented by an extracted text record, stored in the database server.

3) Knowledge system development

The development of the Knowledge system is based on

semantic, conceptual analysis of the learning material. For this purpose, an ontology which defines and models the concepts of the educational portal is developed. In order to facilitate the process of the ontology development, automated production of the ontology using open source tools has been chosen. Tools like OntoLearn [26], OntoLT [27], or TexttoOnto [28], offer guidance and help to develop ontologies from specific data sets, even from text data. All the above tools share the same philosophy of design; concepts definition is the first step, followed by concept modeling through concept recording to a database and

subsequent identification of possible relationships between the concepts. The final ontology is developed by forming the identified relationships as classes with specific attributes.

After collecting the extracted data set from the database server, automatic text analysis via open source language technology tools for concepts definition is performed. Specifically, the GATE tool is used [29], which is a general architecture for knowledge mining that supports many languages. Combined with open source lexicon databases, this architecture performs linguistic and syntax analysis of the text documents. The extracted data are separated in objectives, verbs, adjectives and adverbs because the text analysis obeys to grammatical and syntactical rules. Then the data is annotated and tagged using XML. As a result of this step, a semantic correlation between the content information is produced, which leads to the definition of concepts. The emerging concepts from each web page are recorded and stored in the database server. Yet, machine learning algorithms are applied to distinguish the possible relationships among the concepts. These relationships are grouped into ontology classes resulting to the semantically enrichment of the learning content.

After the learning content has been semantically modeled, the evaluation process follows. When a mobile user visits a web page of the LMS the knowledge system initially checks which semantic data that page contains. Then, the knowledge system searches for semantic data related to the remaining pages of the educational portal and loads the user’s profile data from the database server. Respectively to the semantic data relationships (by concept, instance, similarity, subclass of, instance of, relation), the system recommends links to pages with similar content.

This approach has the additional advantage that feedback information between the knowledge system and the process of concepts definition can be provided. The ontology production tools [28] store indicators for each object in the ontology’s model, revealing the data from which they are emanated and allowing the system to understand why a certain concept or a relation has been created. When changes occur in the text of a particular web page, the system can discover these changes and redefine the concepts or create new ones, without reprocessing all the pages. Then, the ontology can easily be updated, hence leading to increased efficiency of the overall system.

4) Link suggestion to mobile user

The pages suggested by the system may not be directly

connected to those the user has visited. This is the case that the result is even more unexpected and more useful for him. Then, it is critical to present the outcome of the knowledge system, taking into account the limited screen size of mobile devices. Using the basic principles of human computer interaction, a usable mobile application has been designed, which presents the suggested links and guide the mobile learner to them with just one click.

265

Figure 2. Knowledge system description

IV. CONCLUSION In this paper a methodology for improving the mobile

learners’ web experience by personalizing the presentation of educational content is presented. Using web content semantics and user navigational information, our approach aims to provide usable browsing based on page recommendations that are consistent with preferred educational content. The innovation of this approach lies firstly on identifying the mobile users who visit the LMS and their navigational patterns, while the LMS’s data is semantically enriched with ontology concepts.

Contrary to other studies that have been mentioned in the related work (section II), in the proposed architecture the recommendation factors, usage-based records and educational content semantics are defined and their quality is measured in correlation to the users’ navigational patterns. Moreover, there are open source content recommendation systems [30, 31, 32] which make building and maintaining web sites faster and more convenient. Such systems store the actual content in a database and they can then automatically pull the content out and show it on the appropriate pages based on rules that administrators set up in advance. The ways a learning portal can be organized and the types of rules used, depend on how structurally flexible the recommendation system is. Popular open source recommendation systems, like Drupal [30], Plone [31], or Joomla [32], can support a wide variety of site structures and site administrators can define properly detailed rules as to

what content should be displayed to users. Especially, Drupal's taxonomy system enables developers to associate a node with one or many descriptive terms and to create a classification system for content that suits to information architecture needs. However, Drupal’s power comes with complexities when used in an already developed web site or learning management system. Although Drupal’s approach can support a wide variety of site structures when used for site design from scratch, it is rather difficult for Drupal system to understand the structure and to configure node relationships of an LMS built by other tools or different philosophy of design. Drupal’s flexibility is based on thinking through the best way to accomplish what you want before diving in [34]. Moreover, when using Drupal, the administrative screens for configuring a site have a huge number of options and settings, making them harder to interpret [33, 34].

In contrast to the above systems, our goal is neither mobile learner’s sole direction in the educational portal, nor adapting the site’s structure, but providing additional and specific information to the mobile learner, directly related to his educational interests. The mobile learner has nothing else to do than visiting the portal, while the toil of the site’s administrator remains insignificant. As the limited screen size forces mobile learners to a line-based navigation, the proposed solution allows navigating towards their interests with less effort and frustration, and, more important, by saving browsing time. Additionally, since users’ data collection refers only to the specific educational portal, privacy issues, which are a main concern in the majority of personalization methods, are avoided. It is therefore obvious that the proposed methodology is focused on the benefits of personalization methods, while minimizing the costs. On the other hand, the proposed approach has some limitations. The system is designed for educational portals that include mainly text information in their web pages. In portals where text information is scarce, the proposed methodology is likely to fail as the semantic representation and the definition of concepts will suffer from confusions.

REFERENCES [1] Tombros A., Ruthven I, Joemon Jose, “Searchers' criteria for

assessing web pages” Proc. of the 26th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 385-386. Toronto, Canada, 2003.

[2] Barry C.L., Schamber L, “User's criteria for relevance evaluation: a cross-situational comparison", Information Processing and Management, Vol. 34 No.2-3, pp.219-36, 1998.

[3] Nielsen, J., “Designing Web Usability: The Practice of Simplicity.” New Riders Publishing, Indianapolis, ISBN 1-56205-810-X, 2000.

[4] O’Malley C., Vavoula G., Glew J., Taylor J., Sharples M. & Lefrere P. “Guidelines for Learning, Teaching and Tutoring in a Mobile Environment”, MOBIlearn Project Deliverable D4.1, 2003.

[5] Trifonova A., Ronchetti M. “Where is Mobile Learning Going?” Proceedings of the World Conference on E-learning in Corporate, Government, Healthcare, & Higher Education , Phoenix, Arizona, USA, pp. 1794-1801, 2003.

[6] Klopfer E., Squire K. & Jenkins H. “Environmental Detectives: PDAs as a window into a virtual simulated world.” IEEE International Workshop on Wireless and Mobile Technologies in Education. Vaxjo, Sweden: IEEE Computer Society, 2002, pp 95-98.

266

[7] M. Sharples, "The Design of Personal Mobile Technologies for Lifelong Learning," Computers & Education, vol. 34, pp 177-193, 2000.

[8] Albers M. J, Kim, L “User Web browsing characteristics using palm handheld for Information Retrieval”, IEEE Technology & Teamwork, pp. 125-135, 2000.

[9] Jones M, Marsden G., Mohd Nasir N, Boone K, Buchanan G., “Improving web interaction in small screen displays”. Proceedings of Web 8 conference, pp. 51-59, Toronto, Canada, 1999.

[10] Dunlop M. D, Davidson N, "Visual information seeking on PDA top devices”, Proceedings of BCS HCI 2000, Sunderland, UK, Volume II, 2000, pp. 19-20, 2000.

[11] H.M. Chen and P. Mohapatra, “A Novel Navigation and Transmission Technique for Mobile Handheld Devices”, Technical Report, CSE-2003-1, UCDAVIS, 2003.

[12] O. Buyukkokten, H. Garcia-Molina, A. Paepcke “Text Summarization of Web Pages on Handheld Devices”, Proc. Workshop on Automatic Summarization, Pittsburgh, USA, 2001.

[13] Quist G, Goldstein M. “Toward an Improved Readability on Mobile Devices: Evaluating Adaptive Rapid Serial Visual Presentation”, Proc. of the 4th International Symposium (Mobile HCI’2002), pp 240 –255, Pisa, Italy, 2002.

[14] G. Lorho, J. Hiipakka and J. Marila, “Structured Menu Presentation Using Spatial Sound Separation”, Proc. of the 4th International Symposium (Mobile HCI’2002), LNCS 2411, pp 419–424, Pisa, Italy, 2002.

[15] F.J. Gonzalez-Castano, L. Anido-Rifon and E. Costa-Montenegro, “A New Transcoding Technique for PDA Browsers Based on Content Hierarchy”, Proc. of the 4th International Symposium (Mobile HCI’2002), LNCS 2411, pp 69–80, Italy, 2002.

[16] K. Henricksen and J. Indulska, “Adapting the Web Interface, An Adaptive Web Browser”, Proceedings of the Second Australasian User Interface Conference, (AUIC 2001), pp 21-28, Queensland, Australia, 2001.

[17] Hinz M., Fiala Z., Wehner F., “Personalization-Based Optimization of Web Interfaces for Mobile Devices”, Proc. Mobile HCI’2004 – 6th International Symposium, pp 204–215, Glasgow, UK,2004.

[18] Fu Yongjian, Ming-Yi Shih, “A Framework for Personal Web Usage Mining.” International Conference on Internet Computing (IC’2002), pp 595-600, Las Vegas, USA, 2002

[19] Zhou Baoyao, Siu Cheung Hui, Alvis C. M. Fong, “Web Usage Mining for Semantic Web Personalization.” Workshop on Personalization on the Semantic Web (PerSWeb’05),pp 67-76,Edinburgh, UK, 2005.

[20] Mobasher, Bamshad, Honghua Dai, Tao Luo, Yuqing Sun, Jiang Zhu, “Integrating Web Usage and Content Mining for More Effective Personalization.”, Proc. of the First International Conference on Electronic Commerce and Web Technologies. London: Springer-Verlag, 2000, pp 165 - 176.

[21] Mobasher, Bamshad, Robert Cooley, Jaideep Srivastava, “Automatic personalization based on Web usage mining.” Communications of the ACM, 2000, Volume 43, Issue 8, pp 142 - 151.

[22] Thor Andreas, Erhard Rahm, “AWESOME: a data warehouse-based system for adaptive website recommendations.”, Proceedings of the Thirtieth international conference on Very large data bases - Volume 30. VLDB Endowment, pp 384-395, 2004.

[23] Eirinaki M., Vazirgiannis M., Varlamis I., “SEWeP: using site semantics and a taxonomy to enhance the Web personalization process.” Proc. of the ninth international conference on Knowledge discovery and data mining, (SIGKDD), ACM, pp 99-108, 2003.

[24] Michalski R.S, Kaufman K.A., “Data Mining and Knowledge Discovery: A Review of Issues and a Multistrategy Approach,” In Machine Learning and Data Mining: Methods and Applications, Michalski, R.S., Bratko, I. and Kubat, M. (eds.), London, John Wiley & Sons, pp. 71-112, 1998.

[25] Cooley Robert, Bamshad Mobasher, Jaideep Srivastava “Data preparation for mining world wide web browsing patterns.” Knowledge & Information Systems, Springer-Verlag, 1999, pp.5-32.

[26] Paul Buitelaar, P. Cimiano, B. Magnini “Ontology Learning from Text: Methods, Applications and Evaluation”, IOS Press, 2005.

[27] OntoLT, http://olp.dfki.de/OntoLT/OntoLT.htm. [28] Text2Onto. http://ontoware.org/projects/text2onto/ [29] Natural Language Processing Sheffield University. GATE Home,

http://gate.ac.uk/. [30] Drupal, http://drupal.org/ [31] Plone CMS: Open Source Content Management, http://plone.org/ [32] Joomla, http://www.joomla.org/ [33] The Content Management Comparison , http://www.cmsmatrix.org/ [34] Idealware report comparing content managing systems,

http://www.idealware.org/

267