Juan Manuel Adán Coello

An agent for Web information dissemination based on a genetic algorithm

We present and evaluate the architecture of a personal agent that mines Web information sources and retrieves documents according to users interests. The agent represents and retrieves documents using classical information retrieval techniques and uses users feedback and genetic algorithms to learn and adapt to changes in users interests. The agent was customized to help professionals that need to be up to date with the Brazilian Tax Legislation, stored at the Brazilian Federal Revenue site. Experiments discussed on the paper shown that the agent has good precision and recall rates when retrieving documents, and are also able to successfully detect changes in users interests.

A Memory-based Collaborative Filtering Algorithm for Recommending Semantic Web Services

This paper focuses on the construction of collaborative filtering (CF) recommender systems for Web services. The main contribution of the proposed approach is to reduce the problems caused by sparse rating data - one of the main shortcomings of memory-base CF algorithms - using semantic markup of Web services. In the presented algorithm, the similarity between users is computed using the Pearson correlation coefficient, extended to consider also the ratings of users for similarity services. Likewise, to predict the rating a user would give to a target service, the algorithm considers the ratings of neighbor users for the target service and also for similar services. Experiments conducted to evaluate the algorithm show that our approach has a significant impact on the accuracy of the algorithm, particularly when rating data are sparse.

Automatic composition of semantic web services using A-Teams with genetic agents

Automation of Web service composition is one of the most important problems in Web service research area. There are numerous methods to achieve Web service composition. This paper describes the use of Asynchronous Teams (A-Teams) algorithm with genetic agents to compose semantic Web services. Specific agents realize the composition of sequential, parallel and synchronization control flow patterns. Other agents, based on genetic algorithms, perform the crossover and mutation over these patterns. The composition is described through semantic logic rules that take into account the input and output parameters obtained from OWL-S files. The quality of the composition is also evaluated. A system was implemented and typical test scenarios are also presented.

Improving dynamic Web service selection in WS-BPEL using SPARQL

The dynamic selection of Web services must be able to adapt to a changing environment as services are permanently inserted, removed and modified. However, most composition languages do not support such dynamic selection, or support it in a restrict sense, as in WS-BPEL. In this work we employ available technologies, especially the WS-BPEL and SPARQL languages, in order to develop a model for the dynamic selection of Web services. In this model the user expresses the service requirements in the form of a query that will be resolved at runtime. This guarantees the adaption of the system - in a changing environment - without requiring to refactor the entire composition. The restrictions of such a system are listed. An implementation of the model and a case study are reported as well.

Engineering capstone project an individual approach

This article presents a capstone project (CP) approach based on a two-semester effort developed individually to build up and reinforce skills fostered during a computer-engineering curriculum. The CP result is an artifact to be inserted in a context, which is developed, under guidance and monitoring. Besides the student, there are other three roles involved: the advisor for methodological issues; the co-advisor, for technical issues; and the client, to be the source for context knowledge. The student is responsible by planning, developing and managing the project, reporting it in a monograph that is presented in a public defense together with the artifact. The artifact must present a level of complexity that allows measuring some considered major skills. Student and co-advisor choose a methodology for the development effort, whose activities are planned and scheduled. The student should identify open problems in the considered context. One of these problems is chosen as an objective, which should be assessed. Assessment is considered an important issue in the development process. Results are promising: artifacts became real solutions, some even patented; students began graduation because of their CPs or were invited to work because of the acquired knowledge and skills; and students succeed with high self esteem and better communication skills.

SEA: Um Sistema Emissor de Alertas para Fóruns de Discussão, Baseado na Categorização de Mensagens e Avaliação pelos Pares

Webgrade is an on-line system for managing and publishing student grades on web. Webgrade provides: (1) to the teachers, features for managing grade plans, publishing student grades, and sending messages to the students by email; (2) to the students, access to their grades and number of absences to the classes, in a private and restrict page on the internet. All interactions between Webgrade and users are done by a web browser. Webgrade was developed with open-source technology and can be freely distributed under GPL.Resumo: Este artigo relata uma experiencia de inclusao digital (ID) do Programa de Extensao Onda Digital com 40 jovens carentes. O curso de Informatica Basica de 36 horas fez parte do programa educacional da ONG Eletrocooperativa. Este programa visa a formacao musical, melhoria das condicoes de trabalho e de cidadania de jovens afro-descendentes brasileiros. A experiencia de ID contou com a participacao de alunos do curso de graduacao em Ciencia da Computacao da UFBA, que atuaram como instrutores. O gerenciamento do curso, a elaboracao das aulas e do material didatico foram feitos com a utilizacao de ferramentas colaborativas livres. O proprio conteudo do curso de informatica foi baseado em solucoes livres, dentro do sistema operacional GNU/Linux. O resultado foi encorajador: dos 22 alunos que concluiram o curso com sucesso, 13 ja estao empregados em menos de 6 meses e 2 decidiram matricular-se em curso pre-vestibular. Acreditamos que iniciativas como esta despertam nos jovens a motivacao de engajar-se na universidade ou no mercado de trabalho. Abstract: This paper describes a digital inclusion (DI) experience of the OndaDigital Extension Program with 40 deprived teenagers. The IntroductoryInformatics Course had a total of 36 hours and was part of the EducationalProgram from the NGO Eletrocooperativa. Their program focuses on musicalformation, improvement of work condition and citizenship of afro-descendantsBrazilian youngsters. This DI experience counted with the participation ofstudents from the Computer Science undergraduate course of UFBA asmonitors and instructors. The management of the course, preparation oflessons and pedagogical material was done using a collaborative tool basedon free software. The course content was also based on free softwareapplications using GNU/Linux operating system. The results were veryencouraging: from the 22 students that successfully finished the course, 13found a job in less than six months and 2 decided to take preparatory coursesfor entering university. We believe that initiatives like this one motivateteenagers to aim for the university or enter the job market.

Mining heuristics for scheduling distributed hard real-time tasks

We present a method for producing heuristics to direct the search for solutions in task allocation and scheduling problems. The problems considered consist of allocating and scheduling tasks with precedence and hard real-time constraints into distributed systems. Heuristics are produced in two steps. In the first example of promising and unpromising search, states are extracted from the search trees of previously solved problems. In the second, using the extracted examples, the C4.5 algorithm induces classifiers that label search nodes as promising or unpromising. We conclude discussing the results of experiments that compare the success ratio and number of nodes visited to solve problems when a search algorithm chooses the next node to be examined using a random heuristic and when it uses the induced classifiers.

Induction of search control rules for hard real-time task scheduling

The determination of heuristics control rules to direct search algorithms is usually an ad hoc domain dependent process, where complexity increases with the number of factors to take into account. In this paper, we presented a method for learning control rules using tree induction algorithms. The approach was applied to construct a scheduler for periodic tasks with timing and precedence constraints. An experimental evaluation shows that the scheduler is able to learn effective search control rules from previous experiences, improving its performance.