Indexa o e recupera o de teses e disserta es por meio de sintagmas nominais

Introduction: Discusses the use of noun phrases in the automatic indexing process of theses and dissertations deposited in the UFPE Digital Library of Theses and Dissertations (BDTD-UFPE), on the assumption that noun phrases consist of a better knowledge unit for indexing and information retrieval that individual words, allowing an adequate response to the users information need when searching for information. It presentes the state of the art of noun phrases and their automatic extraction process, as well as its applicability in automatic indexing and information retrieval.Method: Based on text analysis tool (OGMA), analyses the applicability of the extraction of noun phrases in automatic indexing and information retrieval of thesis and dissertations in the context of BDTD-UFPE. Applied to abstracts from Law, Computer and Nutrition thesis and dissertations, the variables could be observed, allowing the research team assess the extraction of noun phrases using: the percentage of accuracy of relevant noun phrases; the error rate extract strings that are not noun phrases, and; the percentage of non relevant noun phrases extracted.Results: The process of extracting noun phrases by OGMA showed different performances for each graduate program, with better performance (better accuracy rate) for abstracts from Law Thesis and Dissertations, followed by Computer and Nutrition ones. This performance difference can be partly explained by the different nature of technical terms presented in the abstracts.Conclusions: It concludes that although there are limitations in the available tools, the application of automated methods of extraction and indexing by noun phrases appears to be quite promising, since the noun phrases are configured as best descriptors and access to documents, eliminating the problems caused by synonymy and polysemy of isolated words.