Conversing learning: applying the wisdom of crowds to assist never ending learning tasks

• Ano de defesa: 2013
• Autor(a) principal: Pedro, Saulo Domingos de Souza
• Orientador(a): Hruschka Júnior, Estevam Rafael
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC
• Palavras-chave em Inglês: Never ending learning; Conversing learning; Social Web
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/14524

In Machine Learning systems we often apply techniques are often applied to learn about real world problems behavior from data. Traditionally, such data comes from instances of datasets (that represents the target problem) which we want to learn from. This approach has been broadly used in many different application domains such as recommending systems, meteorological prediction, medical diagnosis, etc. The recent years of quick development of communications technology, that made the Internet faster and available, made possible the acquisition of information to feed a growing number of Machine Learning applications and, in addition, brought light to the use of human computation and crowdsourcing approaches commonly applied to problems that are easy for human but difficult for computers. Thus, the Social Web has been the focus of many research in Artificial Intelligence and Machine Learning. In this work we want to show how we can take advantage from the Social Web to add value to Machine Learning systems which can actively and autonomously ask for web users help to improve learning performance. This work proposes a model of learning called Conversing Learning that is is based on both, Active Learning and Interactive Learning, and is intended to allow machines to convert its knowledge base into human understandable content and then, actively and autonomously ask people (Web users) to take part into the knowledge acquisition (and labeling) process. The work presents how to apply Conversing Learning tasks to assist Machine Learning tasks, and discusses the success of experiments exploring the subtleties of this model.