Indexação automática e ontologias: identificação dos contributos convergentes na ciência da informação


Automatic and traditional indexing has always been a concern of the area of Information Science (IS). The lack of consistency in human indexing, and the lack of semantics in the automatic are regarded as major disadvantages. One possibility of providing computational systems with greater inference power is the use of ontologies. This proposal aims to identify and analyse studies in the area of IS that address the contributions of ontologies in automatic indexing. It is intended to (i) identify the scientific works, in the scientific journal databases of Library & Information Science, and Information Science & Technology Abstracts that approach this subject, its temporal and geographical distribution; (ii) identify and describe the centrality of the subject approach to the two concepts (automatic indexing and ontologies), and the methodological approach of the respective articles; (iii) identify the contributions present in the articles that make up the corpus regarding the potentialities of the joint use of the two concepts. It was based on an exploratory study based on a systematic review of the literature. The results show the contributions of the ontologies in the automatic indexing, such as: (i) disambiguation of homographs and ambiguous terms; (ii) greater ability to integrate semantic relationships in an automated way; (iii) a navigation and expansion of queries through semantic relationships; (iv) a more accurate and exhaustive retrieval of information. We conclude that the development of systems that use ontologies in its full potential in automatic indexing seek to overcome their lack of semantic capacity. Despite the promising results in this regard, it may be still a little early to speak of effective semantic indexing. © 2017, Brazilian Institute for Information in Science and Technology. All rights reserved.