THE DATA COLLECTION MODULE

OF THE AGENT BASED SEARCH SYSTEM (ABSS)

R. Baran, A. Zeja, T. Orzechowski, A. Dziech and M.Lutwin

Department of Telecommunications,AGH University of Science and Technology

al. Mickiewicza 30, 30-059 Krakow, Poland

Keywords: JADE and JADE-LEAP platform, FIPA, OAI-PMH protocol.

Abstract: The main purpose of this paper is analysis and description of the Data Collection Module (DCM) that is the

main unit of the ABSS system. The aim of the DCM module is to process the metadata for the learning

resources and preparing them for the search algorithms. The components of the DCM module are analysed

with respect to technologies and standards used to their implementation. Advantages of proposed solutions

are also discussed. The subject of the paper is related to the Calibrate (Calibrating eLearning in Schools)

project.

1 INTRODUCTION

Subject of this paper is related to the CALIBRATE

project (Calibrate). The CALIBRATE (Calibrating

eLearning in Schools) is co-ordinated by European

Schoolnet (EUN) and supported by the European

Commission’s Information Society Technologies

Programme(IST). The key aim of this project is to

build the system to support the collaborative use and

exchange of learning resources in schools by

allowing teachers to access resources in a federation

of learning repositories. The ABSS system creates

e-learning platform for collaborative searching. The

Data Collection Module is one of the most important

parts of the ABSS system.

2 THE ABSS SYSTEM

The Agent Based Search System has advantages of

the JADE platform which comply with the latest

FIPA 2000 standard (FIPA). Implementation of the

JADE multi-agent platform is fully based on JAVA

technology and is distributed in Open Source under

LPGL license. The latest version of JADE has been

released on 14th March 2006 (JADE 3.4) (JADE).

The JADE platform interconnects hosts of the

federation - Learning Management Systems (LMSs)

and particular elements of the ABSS system

(especially, the main ABSS server) into the one

virtual machine. Each LMS creates a peripheral

container. Peripheral containers are then joined with

the main container of the Agent Platform. The host

of the main container is the ABSS server.

Containers of the JADE platform across the

ABSS system are illustrated in Fig. 1.

Figure 1: The JADE platform of the ABSS system.

The JADE platform creates an environment for

a group of various software agents, dedicated to

perform tasks for effective search. Tasks of the

ABSS agents can be divided into three main

functional categories:

 Metadata harvesting, analysis and storing;

 Management of user profiles;

 Search.

451

Baran R., Dziech A., Orzechowski T., Dziech A. and Lutwin M. (2007).

THE DATA COLLECTION MODULE OF THE AGENT BASED SEARCH SYSTEM (ABSS).

In Proceedings of the Third International Conference on Web Information Systems and Technologies - Internet Technology, pages 451-454

DOI: 10.5220/0001290604510454

 SciTePress

Search services implemented by agents must be

provided to users. There are two types of users

expected:

 End users, searching by a web browser;

 Developers, that use a search service API.

Web Services (WebS) and the Enterprise Java

Beans with JBoss Application Server (JBoss) were

used to implement search services. The JBoss

Application Server is considered as the best open-

source implementation of the Java 2 Enterprise

Edition technology. Web Services are dedicated to

support machine-to-machine interactions over

a network. They provide interoperability between

various software applications running on different

platforms.

Information processed by the Data Collection

Module is stored in databases under control of the

MySQL DBMS system. Two groups of databases

were used:

 Metadata - designated to hold metadata

collected from the LMSs;

 Users accounts - prepared for information

about profiles of registered users.

3 DATA COLLECTION MODULE

The Data Collection Module (DCM) is a set of the

following agents:

 Collector Agent (CA) – a mobile agent

(multiple implementations);

 Content Management Agent (CMA) – a static

agent (one implementation).

The Content Management Agent resides on the

main container and maintains information of all

connected LMSs. The CMA is responsible for

processing, arranging and storing the metadata that

describe the Learning Objects (LOs) - learning

resources provided by the federation. The Collector

Agent is used to gather metadata from the

repositories.

The Collector Agent migrates to the LMS when

registration process of peripheral container is

succeeded. Peripheral containers are established as

results of specialised services of ABSS-LMS

Component. Services of ABSS-LMS element can be

added and supported by the Spark-Core libraries.

Metadata repositories are accessed via the Open

Archive Initiative Protocol for Metadata Harvesting

(OAI-PMH).

Architecture of the Data Collection Module is

depicted in Fig. 2.

Figure 2: Architecture of the Data Collection Module.

4 ABSS-LMS COMPONENT

Connection between the peripheral and the main

containers and communication between CA and

CMA agents are based on the modified JICP

protocol of the JADE-LEAP platform (JadeLEAP).

Communication must be secure and stable. The

SSL Java protocol and mutual authentication are

used to ensure security of the system. In addition, a

set of complex guard threads is used to perform

control of data flow between peripheral and main

containers.

The ABSS-LMS Component includes:

 JadeLeap – the JADE-LEAP implementation;

 Modified JICP protocol:

 JICPSConnectionABSS;

 JICPSPeerABSS;

 LEAPIMTPManagerABSS;

 Set of specialised procedures (including guard

threads)

 Main configuration file with parameters of

ABSS-LMS Component (lms.properties).

There are two simple and recommended ways of

starting the ABSS-LMS component and reading the

lms.properties file:

 As java applet;

 As java servlet.

Parameters of the lms.properties file are divided into

six categories. The most important are

communication parameters:

WEBIST 2007 - International Conference on Web Information Systems and Technologies

452

 Connection parameters – giving numbers of

ports used to communicate CA and CMA;

 lms.abss_host;

 lms.abss_port;

 lms.local_server_port;

 lms.local_client_port;

 SSL/TLS parameters – indicating paths to the

digital certificates and giving a password:

 lms.keystore_file;

 lms.truststore_file;

 lms.authentication_phrase

 NAT parameters – giving IP address of the

NAT box:

 lms.public_nat_ip.

Communication between the peripheral and the main

containers requires two open ports on the LMS

hosts:

 One public server port;

 One client port.

Connections between the containers, during and

after the registration process, are depicted in Fig. 3.

Figure 3: Connections between the containers.

5 COLLECTOR AGENT

The Collector Agent is a mobile agent designed to

monitor metadata repositories of LMSs.

The Collector Agent is designated to:

 Report a complete information about new,

registered LMS and their repository;

 Monitor content of the repository and report

changes – new item was added, item was modified

or deleted;

 Take a detailed information about added,

modified or deleted items;

 Initially process gathered metadata.

Metadata are gathered using the Open Archive

Initiative Protocol for Metadata Harvesting (OAI-

PMH). The Collector Agent getting an instance of

local (LMS) implementation of OAI-PMH protocol.

The ListRecords method is used to harvest the

metadata. Responses of the ListRecords method,

given in a XML format, are conveyed to the parsing

module. The DOMparser from Xerces2 Java Parser

2.8.1 (Xerces2) has been chosen as a default parser

class of the Collector Agent.

The DOM models of each single record are

filtered and processed to select the required

information. These models are also used to check the

profile of harvested records. The profile should

match arrangements

of Celebrate Application

Profile.

The Collector Agent provides a special class,

called the MetaLOM, which properties refer to the

particular categories of required data. The

MetaLOM object is serialized and transmitted to the

CMA.

6 CONTENT MANAGEMENT

AGENT

The Content Manager Agent is a supervisor of the

data collection process. The main tasks of the CMA

agent are as follows:

 Secure registration of the peripheral container;

 Creation and transfer of the Collector Agent;

 Management of the CA behaviour;

 Receiving the MetaLOM objects;

 Processing the harvested metadata and storing

them in especially designed database, called the

DB_LOMS database.

The CMA co-operates with three other agents of

the main container:

 AMS – Agent Management System;

 DF – Directory Facilitator;

 Converter Agent.

The AMS agent of the ABSS system holds the

parameters of all registered peripheral containers

and working Collector Agents.

The Directory Facilitator agent of the ABSS

system indicates registered Converter Agents and

their services. This way, it gives possibility of

calling the appropriate service of selected Converter

Agent.

The Converter Agents are designed to prepare

the metadata for a particular search method. These

agents are specialized to operate in one selected

language.

THE DATA COLLECTION MODULE OF THE AGENT BASED SEARCH SYSTEM (ABSS)

453

7 CA-CMA CO-OPERATION

Messages between agents of the Data Collection

Module are exchanged in asynchronous mode. Every

message transferred in the CA-CMA system is

signed by the appropriate ontology. A set of valid

ontologies is defined by the LMSOntology class.

When migration of the CA is successfully done,

the “MOVE_COMPLETED” message is received by

the CMA. Next, the “START_COLLECT”

command is send to force the CA to start the

initialisation process - the ListRecords method is

used to collect

a complete content of the repository. Obtained

records are parsed and the appropriate MetaLOM

objects are signed by the “LOM_ADDED”

ontology name. Collection of transferred MetaLOMs

is followed by the “END_COLLECT” message.

From this moment, the CA agent is used to harvest

only these records which have been added, modified

or deleted.

8 CONCLUSIONS

The Data Collection Module plays a major role in

the ABSS system. The DCM module is based on the

JADE platform. The JADE platform, according to

the FIPA specifications, makes interoperability and

interconnections between particular elements of the

Data Collection Module easy and secure.

The mobile Collector Agents can be also treated as

significant advantage of the proposed solution. The

Data Collection Module and the other components

of the ABSS system, are implemented in JAVA

language, what makes the system open for the

different platforms.

The ABSS system described in this paper is in

the realization stage. The DCM module is working

with two test repositories. However, the ABSS-LMS

Component is ready to be used to connect next

machines to the federation. Specialized services

guarantee that communication is stable. Metadata

harvested from the repositories are parsed and stored

in DB_LOMS database. Stored information is

processed by the Converter Agents. Actually, the

search algorithms are developed. Algorithms are

implemented as new, dedicated to this purpose,

agents.

The main advantage of the ABSS system is an

open, agent-based and distributed architecture.

Provided researches show that such a system can be

efficient. More comprehensive conclusions can be

formulated when the ABSS system start to work

with federation of authentic repositories.

ACKNOWLEDGEMENTS

The work presented in this paper is supported by the

European Commission under the Information

Society Technologies (IST) program – as part of the

CALIBRATE project, contract IST-28025.

REFERENCES

Calibrate:http://calibrate.eun.org/ww/en/pub/calibrate_proj

ect/home_page.htm

EUN. http://www.eun.org/portal/index.htm

IST:http://europa.eu.int/information_society/index_en.htm

ABSS. http://shalom.kt.agh.edu.pl

FIPA. http://www.fipa.org

JADE. http://jade.tilab.com

LMSs. http://www.eun.org/ww/en/pub/celebrate_help/

glossary.htm

Jboss AS. http://www.jboss.com/products/jbossas

WebS. http://www.w3c.org

LOs.:http://www.uwm.edu/Dept/CIE/AOP/learningobjects

.html

OAI-PMH:http://www.openarchives.org/OAI/openarchive

sprotocol.html

JadeLEAP.:http://jade.tilab.com/doc/LEAPUserGuide.pdf

Spark: http://minor.sourceforge.net/

Xerces2: http://xerces.apache.org/xerces2-j/

WEBIST 2007 - International Conference on Web Information Systems and Technologies

454