2 Marsh, Ghorbani & Bhavsar existing solutions for the information demand problem can be either information retrieval, such as is used in the current

Please download to get full document.

View again

of 35
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Healthcare

Published:

Views: 2 | Pages: 35

Extension: PDF | Download: 0

Share
Related documents
Description
The ACORN Multi-Agent System Stephen Marsh Institute for Information Technology National Research Council Ottawa, ON, K1A 0R6, Canada Ali Ghorbani and Virendra C.
Transcript
The ACORN Multi-Agent System Stephen Marsh Institute for Information Technology National Research Council Ottawa, ON, K1A 0R6, Canada Ali Ghorbani and Virendra C. Bhavsar Faculty of Computer Science University of New Brunswick Fredericton, NB, E3B 5A3, Canada Abstract. ACORN (Agent-based Community Oriented Routing Network) is a distributed multi-agent architecture for the search, distribution and management of information across networks. ACORN utilises the concept of `information as agent' together with an application of Stanley Milgram's Small World Problem (the idea of Six Degrees of Separation) in order to route individual items of information around a network of people and agents. The ACORN ideal is to achieve a state where a web of users is created such that information distribution, queries and search, and browsing behaviour is encapsulated in a single adaptive architecture which learns community behaviour and knowledge in order to route agents to relevant destinations (users). This paper describes the ACORN architecture and its implementation. Weintroduce a novel idea of agent meeting places, or Cafés, to carry out community-based information sharing among mobile agents in ACORN. ACORN is compared with similar work, and evaluations of ACORN for information sharing among mobile agents are described. Applications of ACORN include Business to Business and Business to Consumer based e-commerce solutions, virtual community creation and support systems, peer reviewing systems, and personalized directed information handling. Keywords: Multi-Agent information architectures, autonomous agents, mobile agents, keyphrase matching, multi-agent architecture, community based information handling, e-commerce. 1. Introduction Information is playing an increasingly important role in the networked world. Great changes are taking place in the area of information supply and demand due to the widespread application of computers and the exponential increase of computer networks such as the Internet. The main problems we are facing right now are how to extract relevant, useful, and interesting information from many diverse sources and how to distribute our information to relevant people. Currently, two different technologies are commonly used to address the information demand problem, but fewer systems exist for distributing information. The cfl 2001 Kluwer Academic Publishers. Printed in the Netherlands. fourth-draft.tex; 17/08/2001; 13:51; p.1 2 Marsh, Ghorbani & Bhavsar existing solutions for the information demand problem can be either information retrieval, such as is used in the current crop of Web search engines, or information filtering, for example in services such as SDI (Selective Dissemination of Information). One embodiment of the information filtering technique is the software agent. Software agents exhibit a degree of autonomous behaviour, and attempt to act intelligently on behalf of the user for whom they are working. Agents maintain user interest profiles by updating them based on user feedback. Examples of such systems are the Kasbah framework for e-commerce [7], and matchmaking systems such as Yenta[8,9,10]and ACORN [22]. If we accept that all the Web cannot be indexed [3, 17], either because there are always pages that are new, others are not reachable by any known links, and still others are private (as is the case with many intranets), then the problems discussed above are almost certainly compounded. That is, there are some pages that are not connected to any others (and thus can't be indexed) for one reason or another. An additional problem is that some knowledge cannot be represented on web pages without considerable work this knowledge is social network knowledge, what we call `Who Knows Who Knows What', or WKWKW. Moreover, the contents of the web are constantly changing. All these factors limit the performance of web search engines and information filtering solutions. In this paper we propose an alternative to such solutions. In any reasonably sized network of people, it is fair to assume that the chances of any one individual knowing everything relevant about every other individual in the group are slim. Any one individual, however, may know others in other social or interest groups, so the net result is a collection of interconnected groups of people, linked by `middle men' who know others in different groups. In this sense, we are all `middle men 1 ' in some form or another. In large organisations, too, such collectives of people exist. The Knowledge Management problem in such organisations lies in trying to find out who is doing what, or knows what, in any particular area. Corporate Intranets or employee databases are legitimate attempts to solve this problem, but have tobekept consistently up to date for them to be of any use to the members of the organisation. Moreover, they have to be useful and used. The academic world is a huge web of interconnected practitioners. In this web, information and knowledge sharing is perhaps more open than in other arenas, but still, there are times when it is impossible to 1 Naturally, we don't aim to be exclusive in our terminology. The term `middle person' does not, however, seem to roll easily off the tongue. fourth-draft.tex; 17/08/2001; 13:51; p.2 ACORN 3 know who is doing what in which area. Again, as in Milgram's Small World Problem [23] and see Section 3.2), chains of people exist so that networking, for example at conferences, really can help find people. The ACORN (Agent-based Community Oriented Routing Network) system provides an agent-based peer to peer architecture using community based approaches for information retrieval and provision across networks. It is based on the assumption that a mixture of consumer pull and producer push, coupled with a tight control of information spread, will allow people to keep up-to-date with topics, and will allow the producers of information to get their information in a timely fashion to those who will find it relevant. The agents in the system are autonomous; they make their own decisions about what to do based on information they receive from their creators and from the data they get from other agents in their community. One of ACORN's goals is to facilitate an architecture whereby social networks of computer (ACORN) users could share information based on `people chains.' 1.1. Novelty and Contribution of ACORN ACORN embodies the concept of autonomous mobile information in a peer to peer community of users. Within this community human users are able to dissiminate information in a timely and accurate fashion. In addition, users and agents are able to provide valuable recommendations to non-community members (potentially even non-acorn users). Information within ACORN is able to be rated and filtered by community members and agents, thus providing a valuable means of attaining peer-review of information that may have been previously unrated (and possibly unrateable given current standards of technology). The system uses chains of people and agents to build paths through which mobile agents can disseminate information to like-minded individuals without necessarily requiring those individuals to expressly indicate their interests. In addition, because of its novel information mingling facilities via multiple agent meeting places, or `Cafés,' agents can learn recommendations not only from human users but from each other. ACORN provides an architecture for security which, while conserving user anonymity if necessary and required, nevertheless allows for this peer review and recommendation capacity. The architecture uses the innovative approach of separate agent `brain' and `body,' wherein agent code does not migrate, while agent data does, coupled with a secure agent directory system to provide reliable mobile agent security measures. These measures protect not only the server on which mobile fourth-draft.tex; 17/08/2001; 13:51; p.3 4 Marsh, Ghorbani & Bhavsar agents may be running, but also the mobile agents themselves, both from inadvertent and malicious attacks on their code and/or the data they may be carrying. All of this is managed in a complete, readily available architecture that is implemented in a relatively small footprint (in Java), is extensible to take into account new developments in information sharing and handling, and makes full use of readily available communications and metadata standards such as XML and the Dublin Core [13, 30]. In addition, the system provides for readily adaptable user interfaces because of its adoption of the JSP libraries Potential ACORN Applications ACORN's potential applications include people finding via virtual communities and also via extension to standard web searching approaches, the dynamic building and maintenance of virtual communities of likeminded individuals, and adaptive messaging through intelligent information dessimination. On the e-business front, ACORN can provide facilities such as user preference directed information (i.e., directed advertisements), the dynamic building of business and consumer coalitions, and novel knowledge management tools for larger organizations [19]. Finally, we see ACORN as a suitable replacement for or extension of standard messaging systems, chat and web search architectures. In this context, to coin a phrase, we see ACORN as ` with attitude.' 1.3. Paper Organisation This paper is organized as follows. Section 2 briefly reviews related work. An overview of ACORN is given in Section 3 along with a description of Milgram's Small World Problem. The architecture of ACORN is presented in Section 4. In Section 5, the various types of agents in ACORN are described. We propose a novel idea of Café as a virtual meeting place for mobile agents of ACORN in Section 6, and some relevant methods for information sharing are outlined. We also explore a dynamic clustering method for ACORN Cafés. Salient features of the ACORN implementation, along with performance evaluations of ACORN are described in Section 7. We conclude with pointers to further work and a summary of ACORN's capabilities in Sections 8 and 9. fourth-draft.tex; 17/08/2001; 13:51; p.4 ACORN 5 2. Related Work Community-based navigation and information sharing are not new [12, 16, 18] and interesting twists on the themes exist (or have existed), such as the now extinct Firefly and the vibrant Epinions [14], whichtake the views of a community, aggregate them, match individual interests, and recommend to their members movies, music or whatever that related members are into. These services work, but in a limited sense: people need to go to some non-minimal effort to describe their interests, and they need to keep coming back to recommend new movies, recordings, books, or whatever. Amazon (www.amazon.com) has another way of doing this, by suggesting books to people that others have bought, based on the title they are looking at now. This requires no effort on the part of the `recommender,' and works surprisingly well. The concept of effort is important: people aren't going to use something they have to do extra work for unless they get something more in return. Agents have been used to provide or discern links between people. Foner's Yenta [10] is an example of this. It takes people and their interests and tries to match them with others with similar interests. This is of particular use in a community-building situation, but can also help when looking for information (or possible interested receivers of information you have created); similar systems are described in [15, 16]. 3. Overview ACORN embodies the principle of `Social Knowledge Management,' wherein the knowledge may lie in private yet easily obtainable places such as intranets, or may lie only in the heads of the people in the society. ACORN is a multi-agent based system which uses the concept of `information as agent' to route information around networks (or communities) of people. Information in ACORN's context is anything that can be transported electronically, such as documents, in whole or in part, queries, images, sounds, and so on. The implementation of ACORN referred to in this paper uses mobile agents that are capable of performing both search and distribution of information Motivation Information is hard to work with, particularly if there is a lot of it and it is disorganized, such as on the Web. Current solutions are straining at the seams, and a new paradigm is required to handle the volume and noise. Agents present only a possible solution, but one fourth-draft.tex; 17/08/2001; 13:51; p.5 6 Marsh, Ghorbani & Bhavsar which bears closer investigation. Conceptually, an agent can be sent out onto networks to scour web sites or corporate information sites and databases available to it for information relevant to a search topic it has been given. This is an extension to the `spiders' that the original search indexes used, with directions not to grab every file at a site, but to get relevant pages. Such `searchbots' exist and work relatively well, for all their simplicity. However, humans employ other search strategies, via their communities: asking questions of people you know might result in them going elsewhere to ask another person they know (and so on) resulting in, eventually, an answer coming back to them. This is community-based searching, and requires only that you are a member of a community, and since communities overlap, it is easy to grab information across community boundaries. ACORN uses such a community-oriented approach. The basic questions that ACORN seeks to address are those we live with everyday as members of communities: Who can help me with a specific problem? Who would be interested in this information? Who knows the answer to this question? In addition, we attempt, via ACORN, to answer other questions that might be asked in some situations: Who has read what I wrote, and what did they think of it? How can I control who sees what I write, yet still use an open network to distribute it? Can I build up a team or communityoflike-minded people quickly? 3.2. Milgram's Small World Problem ACORN's primary approach is to use mobile agents to distribute information. ACORN's agents move from person to person, following trails of recommendations. The users the agents visit can recommend other users, thus increasing the length of the chain and the number of people the agent can visit. At the end of its journey, the agent should have learned more about who exists in the community and who is interested in what it has to present. ACORN's behaviour is in fact very close to the effects seen in Stanley Milgram's Small World Problem [23]. In Milgram's experiment, several letters were sent to randomly selected people in the continental US, fourth-draft.tex; 17/08/2001; 13:51; p.6 ACORN 7 with instructions that the letter had to find its way back to a named person (in Boston, since that's where Milgram was). However, the caveat was that the letter could only be forwarded via people whose first name the forwarder knew (in other words, people they knew socially). Taking into accound blind alleys, closed worlds, and the occasioal fluke (for example, one of the randomly selected people just happened to know the postmaster in the town the final recipient lived, so the chain was very short), the average length of referral chains was six people. Hence, the concept of Six Degrees of Separation between any two randomly selected people in the continental US at the time of the experiment. Although things have changed for one thing, there are more people out there certain assumptions about the Small World Problem can still be made. That people still interact with each other, even over large geographic distances, is clear, for example. In addition, people are becoming more, not less, networked [31]. The number of links between people has probably grown, although a new experiment would be required to prove that assumption. Recently, attention has turned naturally to the World Wide Web. The Web is a very connected network of pages, and hence people or organisations. In addition, the Web is very information centric the resulting ties of linked pages are often more akin to citations in scholarly works than social links, although these, too, exist. Recent work in this area has found that the average number of links from one random web page to another is in around 19 [1]. However, the actual shape of the Web may in fact be less straightforward than such simple figures suggest [3], leading to additional problems with the traditional search engines, on top of what they already have to deal with [17]. The Web, however, is almost certainly a Small World [1, 29], or at least a collection of Small Worlds. We conjecture that the people connected through the Web and through networks in general are part of a Small World in themselves, and it is on this assumption that ACORN operates. 4. Architecture Figure 1 shows an overview of the ACORN architecture as it stands today. As can be seen, ACORN is, from the top level, a Client/Server architecture. Clients, the user's interface to the ACORN system, present their user interface via JSP pages and displayed to the user on any standard web browser. Clients provide the ability to create agents, organize received agents, describe a user's interests, and input details fourth-draft.tex; 17/08/2001; 13:51; p.7 8 Marsh, Ghorbani & Bhavsar of other users and their interests. This information is used in the agents to better distribute themselves. Figure 1. ACORN Architectural Overview All community information the Client has is stored in a separate user profile file on the server side. This core contains user information, references to agents received and sent, user preferences, and also information about the people in the user's database. Data about other users is stored as addresses allied with keyphrases and associated weights. The Main Servers in ACORN provide mobility capabilities to agents, and also give ACORN a degree of persistence in information. All Client data is stored (as a client core) on the server side so that, for example, if the client shuts down, incoming agents can still be processed according to user requirements. By the same token, we can provide automatic redirection and recommendations to agents because the server can access the client cores to match agent topics with known user interests in the client core. Because of this structure, the user is able to log on to the ACORN system from any machine with a simple web browser. fourth-draft.tex; 17/08/2001; 13:51; p.8 ACORN 9 Agent migration is achieved by sending agent cores from server to server. Like client cores, the agent core contains all the information the agent needs to accomplish its tasks: addresses of people to visit, user interests, document information and so on. When an agent reaches a new site, the server at that site instantiates a new agent with the core that has just been received, and lets the agent go on its way. At a site, the agent can visit users, and communicate with (mingle with) other agents at a central meeting point (whichwe call a Café) to share addresses and interests with others, thus aiming to extend the list of addresses for people to visit. Recent additions to the architecture are the Directory Server, which provides the ability to track andcontrol agents which are off site, and the Anonymity Server, which provides routines to anonymise InfoAgents before they are sent to the network, and to re-instantiate them as necessary on their return. 5. Agents Client, Server, InfoAgent, Anonymiser, Directory Server In order to provide the infrastructure for the mobile agents to perform their tasks, the architecture uses static agents. These are the Client and the various Servers. In addition, the Cafés, with their managers and blackboard agents, provide for information sharing between agents and persistence of infor
Recommended
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x