Document retrieval system

Abstract

To effectively utilize repositories of data, document retrieval systems act as a means of performing the task of sifting through these repositories to extract documents that meet an individual's information need. The projects' document corpus was drawn from 71 Master of Science in Information Systems and Computer Science project abstracts done at the University of Nairobi, School of Computing and Informatics between the years 2006 and 20IO.The project utilized the vector space model as its basis for document matching and ranking. Based on the gold standard of relevance, these documents were put in document categories that reflected their content. This provided the basis against which recall and precision measures office system accuracy was measured. The system achieved an average precision score of 0.781667 and recall score of 0.833333. Recall scores were mainly affected by the presence of homonyms and homographs while precision scores were affected by synonyms. The study also showed that the presence or absence of a term in a document is what influences the retrieval and ranking of relevant documents in the vector space model, not the size of the document corpus.