A função do resumo na comunicação científica: uma comparação entre resumos tradicionais e estruturados

Abstract

We sought to compare abstracts of scientific articles in the Journal of Documentation (JD) andthe Journal of the American Society for Information Science and Technology (JASIST),evaluating the performance of these abstracts with regard to the representation of relevantcontent from the original texts. The JASIST publishes along with its scientific papers traditionalabstracts. The JD publishes structured abstracts.It is argued that the information society is the society of coded and theoretical knowledge andthat scientific communication is one of the main processes on which this society rests. It is alsoargued that the information retrieval, as an exceptional instrument in the process of scientificcommunication, will advance to the extent that abstracting and indexing advanceconceptually. In particular, the normalization of abstracting must evolve in regard toinstructions and techniques for content selection (and not only for its presentation).We compared 42 traditional abstracts and 42 structured abstracts. First, the abstract wascompared to the scientific article that it condenses. It was used two concepts: textualsuperstructure (a concept made operational by a proposal applied in this study) and thematicstructure (concept analyzed subjectively). The textual superstructure is recognized by a 'basesequence' such as: problem, hypothesis, methodology, results, conclusion. The thematicstructure is the subject of a text. This structure establishes the 'textual similarity' - at themessage level. The textual superstructure divides the text into articulated parts andestablishes the 'textual contiguity' at the structure level (which shows the relationshipsestablished).The analysis with statistical power of approximately 0.79 (using a chi-square with DF=1) foundthat 3/42 or 7.14% of the traditional abstracts provide a satisfactory representation in relationto the original scientific article, and, 14/42 or 33.3% of the structured abstracts provide asatisfactory representation in relation to the original scientific article. The use of textualcontiguity to analyze the original text of the article and its abstract was practical. An abstractstructurally similar to its original provides an effective 'interpretive' representation.