Pedro Azeredo Paixão

Automatic Extraction of Biographical Data: Automatic Collection of Biographical Data on the Web for Public Personalities Pedro Azeredo Paixão

In recent years there has been increasingly the need to take data, organise it and turn it into information. Given that 80% of the available information in the world is in text format, it makes sense to try to extract the information that matters to us. To be able to extract this type of information with quality from a text, it is necessary, first of all, to structure and organise it. To provide search engines or any other system with the ability to understand synonyms of public entities, first we need to be able to acquire the relations between persons and roles/jobs. We created a system that is composed by an automatic information extraction system able to extract new relations along with the entity names and their binding roles; a graphical user interface that allows the manual validation; and a biographical information retrieval system to acquire extra information like article abstracts from wikipedia, article links and photos. Human-validation tests were performed and the results revealed that the main objective, acquiring person-role relations, was well accomplished with accuracy levels greater than 80%.