The application of assisted folksonomy in the construction of Information Science reference corpus

摘要

This paper proposes and discusses the application of assisted folksonomy in the construction of reference corpus of scientific articles of Information Science area. The hypothesis is that the result of the application can guarantee a better quality in the indexing of scientific articles and a better evaluation of the automatic indexing systems through the compiled corpus. The outlined research uses the corpus composed by 60 articles written in Portuguese selected by Souza (2005). The collaborative platform for assisted social indexing of the corpus consisted of the collection management software called Tainacan. The stages of the research involved the configuration and preparation of the collection in Tainacan, the accomplishment of the assisted social indexing by groups of users and analysis of the results of the indexing process. The analysis of the assisted folksonomy consists of comparing what is available in the Subject and tags metadata fields of the articles. The indicators of the quality of indexation were a mean of 28% of the consistency coefficient, 32% of precision, 68% of recall, and 41% of the F measure. The indicators represent good levels of consistency and recall and satisfactory levels of precision and F measure, implying that the use of assisted folksonomy is useful in improving the indexing of the reference corpus.