Proceedings of the 8th Annual Sessions, Sri Lanka Association for Artificial Intelligence

20th December 2011 – Moratuwa


Existing Systems and Approaches for Machine Translation: A Review
B. Hettige, A. S. Karunananda

Abstract-The Machine Translation has been a branchof Natural Language Processing, which comes under the broad area of Artificial Intelligence. Machine Translation system refers to computer software that translates text or voice from one natural language into another with or without human assistance. Worldwide, large number of machine translation systems have been developed using several approaches including humanassisted, rule-based, statistical, example-based, hybrid and agent based techniques. Among others, Statistical machine translation approach is by far the most widelystudied machine translation method in the field of machine translation. The multi-agent approach is a modern approach to handle complexity of the systems in past five years. This paper reviews existing machine translation approaches and systems including existing English to Sinhala machine translation systems.



MASDM – A Multi Agent Solution for Disaster Management
P.V.C.P.Perera, Asoka S.Karunananda

Abstract-In the context of large natural disasters like the Indian Ocean Tsunami in 2004 or Hurricane Katrina in New Orleans in 2005, the topic of disaster management has become a key concern both from a social and political point of view. Problem solving in an environment where a disaster has occurred is difficult and dynamic, requiring rapid decisions before, during and after the event. So a disaster management system should be capable of withstanding uncertainty,ambiguity and incompleteness given the dynamic and evolving nature of disasters. As a solution, a disaster management system based using agent technology has been implemented to provide proper management of resources and responsibilities for dealing with all aspects of a disastrous situation; in particular preparedness, response and recovery in order to lessen the impact of disasters. The system mainly consists of four agent swarms which are: forces swarm, aid store swarm, hospitals swarm and information providers swarm. Each swarm consists of number of agents whose tasks are explicitly defined. Agent interaction is the source of generating the intelligence. Agent communication, coordination and negotiation capabilities provide the ways and means of handling the complex nature of the domain.


Spatial Model of Road Network in Colombo City for Optimum Online Path Finding
S.E. Jeningthas, HUW Ratnayake

Abstract-The Optimum Online Path Finding Systemis an online path guide for bus travelers within the Colombo city. It is designed to show the public, the optimum route that they can take to travel from one place to another since there are several bus routes within the city. The system is implemented using bus route detailsprovided by the National Transport Commission(NTC). The system answers questions from the public regarding bus routes. It shows them possible routes from a source to a destination from where they can make decisions regarding their travel.Optimum path finding algorithm is based on the A* algorithm. The prototype implemented cansuccessfully show graphically, the optimum path with the bus route numbers


Disease Diagnose of Rice Plants using Fuzzy-Expert Hybrid System
T.D.Nuwan Chathuranga, Gehan Anthonys

Abstract-Sustainable agriculture related research is a hot topic in modern world. In Sri Lanka, paddy cultivation is a main income method. Advantages of using these systems are that the output of such systems are very accurate and faster than analyzing the same problem by humans. This paper proposes a web based software system which has the ability to identify pest/s and their damage level to the plant by analyzing the various kinds of inputs such as damage symptoms, size of the pest, damage intensity etc. It is very helpful to identify the condition of the paddy before it becomes critical


Towards Multi-Agent Based Solution for Ontology Alignment
G.M. Rajakaruna, A.S. Karunananda, S. Jayalal

Abstract-Ontology alignment is an ontology mapping technique, which is used for sharing and reusing the content of disparate source of ontologies. Aligning dynamic and distributed ontologies turn out to be a research challenge for many decades. This paper discusses the complexity of the ontology alignment process and presents some major tools that have been developed to support knowledge engineers. Then, we discuss how the agent technology can be used to align dynamic ontologies in distributed environments. Our solution is inspired by how group of people communicate, negotiate, and coordinate to reach common agreements.


Customizing the WEKA for Visual Improvements in the Output of J48 Decision Tree
M.A.S.M. Perera, W.D.S.S. Appuhami, A.N.P.Bandara, D.S.Deegalla

Abstract-The integration of communication and computing has dawned an enriched era of society which feeds on information. Yet, a huge amount of potentially useful and implicit information is locked up in databases as raw data. Data Mining (DM) is an active research area which deals in finding patterns, anomalies and trends in this data and summarizing them with quantitative and qualitative models to predict future aspects. In this paper, we describe the visual improvement of the output produced by the J48 decision tree algorithm in the WEKA toolkit, which is an open source DM toolkit widely used in teaching as well as research. We improved the visual representation of decision trees by coloring nodes according to the class distribution of the instances reached at each node, displaying values of information gain and gain ratio of intermediate nodes and displaying decision rules of leaf nodes accordingly. These improvements enhance the model’s user friendliness and understandability.