CLELIA: a multi-agent system for publishing printed and electronic media

https://doi.org/10.1016/j.eswa.2004.12.029Get rights and content

Abstract

This paper describes a multi-agent system (MAS) for automatic publication of information in both printed and electronic media. The main objective is to have a document ready to be displayed (and printed) at every stage of the workflow. This is accomplished by replacing the traditional human role of page editing by a multi-agent system composed of three different agent roles (section, page and element). Key aspects such as page aesthetics or readability are handled by using artificial intelligence techniques such as neural networks, genetic algorithms and fuzzy logic. The theoretical foundations of this system are discussed and an example of implementation using an external toolkit is described.

Introduction

New perspectives concerning the future of the Internet are founded on an essential idea: if information is to be processed by automatic means in a proper manner, systems need to comprehend to a certain extent (semantically) the information they are dealing with. The current trend in information systems architecture leads to the replacement of the client-server model of today by distributed services in a completely heterogeneous, mobile and distributed environment. This is possible thanks to some common protocols based on SOAP and XML (the so-called ‘Web Services’ technologies). Following this viewpoint, information itself must be expressed and evaluated in the same way by all the participant entities, focusing not just on displaying issues (html, script languages and so forth) but on creating a global semantic structure underlying the information to be displayed. The vortex between these new approaches on application architecture (distributed web services) and information structure (semantic web—Berners-Lee, Hendler, & Lassila, 2001) might be ‘intelligent’ agents.

An ‘intelligent’ agent can be described as a conventional agent (that is, an autonomous entity capable of interacting in a given environment, to manage resources and to communicate with other agents to accomplish an objective), able to emulate features of human reasoning (Von Neumann, 1958) by artificial intelligence techniques such as neural networks, expert systems, fuzzy logic and genetic algorithms. These entities are conceived as nodes dealing with information on behalf of a certain human or an automatic client (another agent) to whom they offer some kind of service. By incrementing the amount of ‘intelligent’ capabilities on these agents, client requests can be expressed intuitively while the underlying operations might increase in complexity offering users almost unlimited functionalities.

The domain of publication is well-suited for the testing of such technologies. Before the information reaches the end user by electronic (push–pull procedures) or paper-based means, it does have to be displayed in a suitable way. It is at this stage where all the breakthroughs on middleware layers might reach a pitfall due to a sequential conception of work and information flows. At the first stage, information is collected through data requests or introduced into the system, producing an XML document. At the second stage this information is either manipulated or used directly to fit a format pattern, such as an XSL file. There is no automatic or simultaneous interaction between both procedures. By blurring the border between obtaining the required information and displaying or printing it under a specified format, we can offer more up-to-date and reactive information, as happens in the system for text-based media described in this paper.

A few years ago, the electronic media began to offer personalized newspapers only to discover shortly afterwards that their clients (or readers) were not using such an appealing idea. This was mainly due to three different reasons:

  • They would not voluntarily make the effort to determine the format and type of information they wanted to receive.

  • The options such a system offered to the clients, provided they were willing to personalize it, were quite frustrating. Usually they would oscillate between choosing sections (for instance to receive only international news and sport sections), or presenting a front page composed of a collection of links based on generic topics or keywords readers had previously selected and pointing to a few static news-item pages (as happens today with syndication software).

  • A more generic reason was that users still prefer the unbeaten visual comfort and ergonomics that traditional paper-based media has to offer, relegating electronic sources to last-minute news and overall outlook on daily issues.

On behalf of this third point, printed and electronic media will be analyzed together in this paper. As for the other two points, what users and systems are longing for is an adaptation to a changing external context (the different readers needs and habits). Electronic papers need to take advantage of their support and be more ‘intelligent’ to fully comprehend user desires far beyond requests by keywords. In addition, they should offer a visual page outfit equal in quality to printed media instead of presenting just a collection of links or static XSL templates where dynamic XML data is embedded. Printed media, on the other hand, while having an excellent visual quality, presents the logic drawbacks of low reactivity and scarce personalization.

Text-based media usually follows a four-step process: the chief redactor for a given section proposes some topics, the journalists write articles on them, the publishing team emplaces these elements on pages (physical, electronic or both) and finally the printing house prints them on paper, on the website or both. These workflow and information-flow models pose some issues regarding information redundancy, underused resources and low compatibility on data formats. However, the main problem remains at the inherent sequenciality of the process: nowadays it is necessary to completely finish one stage before moving on to the next. By automatic publishing, printed media could offer more up-to-date journals by reducing the delay up to several hours from the redaction of an article to its publication. Automatic editing carried out while the journalists are writing would make it possible to publish the document right after it is finished or even incrementally if needed. Decision-making shall prevail with appropriate modeling patterns.

Section snippets

Objective

The main objective for the proposed system is to overcome the sequential workflow for both printed and electronic media as it exists today by creating a multi-agent system (MAS) where information creation and its publication can be simultaneous. This will allow the media to be more reactive, optimize resources consumption and offer the end users more personalized and visually appealing products.

Conceptual architecture

After establishing the main purpose for the multi-agent system and targeted the key problems to deal with, this system is now conceived and classified by using generic concepts in a bottom to top methodology.

CLELIA components and interfaces

This multi-agent system has been conceived as an autonomous component, that is, an independent, interchangeable and interoperable module, which may be integrated in any actual or future publishing software. This could be done either on a web-service premise (as an external functionality offering automatic editing) or in a more conventional bottom-to-top application architecture. At this stage it has been deployed under the latter scheme, as a component box that interacts with the upper and

Example

The example provided here allows the reader to have a specific vision about how the multi-agent system accomplishes the previously described tasks in a realistic problem and a step-by-step approach.

Conclusion

Information systems are moving towards a new perspective focused on distributed systems offering services and a certain understanding of information via semantic web techniques. Nevertheless information itself still needs to be displayed on conventional paper or screen interfaces for human consumption. Merging system and user communications in a common environment would break a sequential workflow that remains in the overall architecture: first obtaining or producing the information in XML,

Acknowledgements

The research work leading to CLELIA was initially developed at the EPF-École d'Ingénieurs (Sceaux, France) as part of a software project called Tangramme (which integrated additional network, GUI and collaborative features). A thorough revision, as well additional research was possible thanks to additional funding and help provided by the IIT-Instituto de Investigación Tecnológica (Madrid, Spain).

The authors wish to express their gratitude to PhD. Monsieur Pierre Grignon, who initially

References (6)

  • T. Berners-Lee et al.

    The semantic web

    Scientific American

    (2001)
  • J.P. Bigus et al.

    ABLE: a toolkit for building multiagent autonomic systems

    IBM Systems Journal

    (2002)
  • C. Bishop et al.

    Neural Networks for Pattern Recognition

    (1996)
There are more references available in the full text version of this article.

Cited by (0)

View full text