Elsevier

Computers in Human Behavior

Volume 51, Part B, October 2015, Pages 1320-1330
Computers in Human Behavior

Creating recommendations on electronic books: A collaborative learning implicit approach

https://doi.org/10.1016/j.chb.2014.10.057Get rights and content

Highlights

  • It is possible to develop more efficient recommender systems that does not depend on users’ explicit ratings.

  • It is possible to determine users’ interest by analyzing and converting their behavior.

  • This approach allows to build a collective web knowledge in a co-learning context.

  • This work allows to analyze the users’ behavior data and convert it into explicit ratings.

  • 75.6% of data obtained by the conversion mechanism is close to optimal values.

Abstract

Recommender systems appear among other reasons with the purpose to improve web information overload and ease information recovery. This kind of systems aid users to find contents in a non-difficult way and with minimal effort. Even though, a great number of these systems performance requires contents to be explicitly rated in order to determine user’s interest. When interacting with electronic books this performance may alter users reading and understanding patterns as they are asked to stop reading and rate the content. Therefore, the analysis of user behavior, preferences and reading background can be considered suitable for a recommender system to build collective web knowledge in a collaborative learning context. This way, recommender system can assist users in finding contents of their interest without explicit rating based on previous constructed knowledge. The goal of this research is to propose an architecture to build a content recommendation platform based on eBook reading user behavior, allowing users to learn about the digital content collaboratively. This platform is formed by web readers’ community that aids members in finding contents of their interest in an automatic way and with minimal effort.

Introduction

Last years, the exponential growing of the information available on the web brought an issue called information overload. Hence, the great data amounts makes difficult to discover, find and classify the most relevant information for each user profile or interests (Zhang, Zhou, & Zhang, 2011). Commonly, users seek for recommendations from another users or media in order to find the most valuable information or products they need (González Crespo et al., 2011, Su and Khoshgoftaar, 2009). Recommender systems are usually employed to deal with information overload on the web as an information recovering and classification technique. They filter the information available on the web and help users to find more interesting and valuable information (Noor and Martinez, 2009, O’Donovan and Smyth, 2005, Taghipour and Kardan, 2008). The most relevant search engines like Google, or online stores like Amazon, have incorporated recommender technologies as part of their services with the purpose to personalize the search results (Verbert et al., 2012).

Despite the major upswing and extensive utilization of these systems, there is a gap in the information feedback process, which is a key part of all the recommendation process that is susceptible of improvement. This paper sustains that recovery, analysis and transformation of user behavior can be used to measure their interest in some determined contents and therefore be able to bring more accurate recommendations to them. Even though, as illustrated in (Claypool, Brown, Le, & Waseda, 2001), the most common solutions and the more prevalent are the ones based on explicit ratings. In the context of eBooks these techniques can alter the user’s regular navigation and reading patterns, because they have to stop and rate the items.

In (Nuñez-Valdéz et al., 2012), it was recently defined a set of implicit parameters on which was performed a comparative analysis that led to the correlations between the actions that a user can perform during an eBook reading time and the explicit ratings given by it on each content. These findings showed that is possible to determine user interest through the analysis and transformation of its behavior. Taking into account these results and, with the implementation of an architecture that contains an algorithm to perform this transformation, recommender systems can be constructed in a more precise manner, based on implicit feedback.

In these times of information overload, emerges the necessity to develop recommender systems which allows discovering users’ interest in a more effective and simple way improving their experience and satisfaction. The possibility of analyzing and studying the users behavior on a social network of electronic books allows us to improve the collaborative learning of its members. The use of recommendation systems allows readers to create and share collective knowledge in an easily and automatically way.

The rest of this paper is structured as follows Section 2 presents the background of recommender systems; Section 3 shows a case study and the architecture proposed; Section 4 presents the evaluation of results; and finally, in Sections 5 Discussion, 6 Future research directions are explained the future research directions and conclusions of this work, respectively.

Section snippets

Background

Recommender systems are tools that aid users to find the information they really need in an easy and efficient manner. These systems helps to optimize the time users employ in searching contents that somehow are harder to find. These contents are selected by recommender systems from a large amount of data that is available on the web and can be any kind, such as books, movies, songs, websites, blogs (González Crespo et al., 2011).

Recommender systems are based on personalized information filtering,

Case of study

One of the main issues of recommender systems is the deficit on the implementation of information feedback mechanisms. The main reason of this deficit in most of the cases takes place because of these mechanisms are based on explicit feedback which can be an inconvenient for users as they usually do not like to rate contents. Hence, if users do not rate contents it is not possible to know the contents of their interest, reason why it cannot be possible either to recommend contents to them by

Evaluation

In this section the proposed architecture is evaluated. To do this, the results obtained from the platform’s implementation are analyzed, focusing on the resultant data from the explicitation process of users’ behavior. These results show a clear sight of the users’ interest on the contents, obtained through the analysis and interpretation of their actions.

The evaluation is done for each user and content within the platform through the comparison of the explicit ratings given by the users to

Discussion

As it has been specified in the study case, the aim of this section is to validate the results obtained from the proposed architecture and, in particular, determine the effectiveness of the User Interactions Converter Algorithm (UICA) with which the behavior of the users of an electronic book social platform is analyzed and converted into a set of values that are considerably close to the explicit feedback.

Analyzing the evaluation of the explicitation results for the behavior of the users (with

Future research directions

Although we have performed a first approach to the use and implementation of recommendation systems based on implicit feedback, this proposal can be refined and extended to bear additional features. Also, it opens different investigation paths to complete and improve the defined methods and tools. Some of these paths are: (1) Implementation of this architecture in other environments, which allow the recommendation of any kind of products. (2) Defining a Domain Specific Language (DSL) that

Conclusion

In this paper we have proposed an architecture for the construction of a content recommendation platform based on the behavior of the users of electronic books in the web, aiming to help the users discover contents of their interest automatically and effortlessly.

The goal was to achieve an approximation to a solution for the explicit feedback in the recommendation systems within an environment of electronic books. This architecture allows to analyze the behavior of the user and convert this

Acknowledgments

This work was performed by the University of Oviedo under Contract No. TSI-020110-2009-137 of the research project eInkPlusPlus. Project co-financed by the Ministry of Industry, Tourism and Commerce under its National Plan for Scientific Research, Development and Technological Innovation.

References (26)

  • H.J. Lee et al.

    MONERS: A news recommender for the mobile web

    Expert Systems with Applications

    (2007)
  • A. Montes-García et al.

    Towards a journalist-based news recommendation system: The Wesomender approach

    Expert Systems with Applications

    (2013)
  • G. Adomavicius et al.

    Incorporating contextual information in recommender systems using a multidimensional approach

    ACM Transactions on Information Systems (TOIS)

    (2005)
  • G. Adomavicius et al.

    Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions

    IEEE Trans. Knowl. Data Eng.

    (2005)
  • M. Balabanović et al.

    Fab: Content-based, collaborative recommendation

    Communications of the ACM

    (1997)
  • M. Claypool et al.

    Inferring user interest

    IEEE Internet Computing

    (2001)
  • González Crespo, R., Sanjuán Martínez, O., Cueva Lovelle, J. M., Pelayo García-Bustelo, B. C., Gayo, J. E. L., & de...
  • H. Kautz et al.

    Referral Web: Combining social networks and collaborative filtering

    Communications of the ACM

    (1997)
  • Kelly, D., & Belkin, N. J. (2001). Reading time, scrolling and interaction: Exploring implicit sources of user...
  • D. Kelly et al.

    Implicit feedback for inferring user preference: A bibliography

    SIGIR Forum

    (2003)
  • G. Linden et al.

    Amazon.com recommendations: Item-to-item collaborative filtering

    Internet Computing, IEEE

    (2003)
  • Noor, S., & Martinez, K. (2009). Using social data as context for making recommendations: An ontology based approach....
  • E.R. Nuñez Valdéz et al.

    Social voting techniques: A comparison of the methods used for explicit feedback in recommendation systems

    International Journal of Interactive Multimedia and Artificial Intelligence

    (2011)
  • Cited by (0)

    1

    Tel.: +34 985182451, +34 985103397; fax: +34 985181986.

    View full text