Enriquecendo um arquivo de autoridade de veículos de publicação com informações extraídas da Web

Abstract

Authority files maintain entity registries and are generally used by digital libraries for elaborating disambiguation tools for author names or titles of publishing venues. An authority file with detailed and consistent information on publication venues allows the improvement of such tools. This work has the objective of enriching an authority file of Computer Science publication venue. The proposal is of obtaining additional information in order to complement this already existing authority archive, by automatically extracting information from web pages, obtained by means of consultations to a research engine. The approach contemplates the steps for submitting consultations, classifying documents and extracting information of relevant documents. The classification of the pages is an important task in this work. Two approaches were implemented and experimentally evaluated: classification based only on content, and classification based on gender and content. The first obtained the best results for page conference. From the relevant pages, we extracted data such as year, edition number and date, in addition to name and abbreviation, seeking an unknown variant in written form. The experiments conducted demonstrate good results in the collection of conference information, allowing us to trace the record of performing the same, with data such as edition year and name change.