An evaluation of the applicability of ranking algorithms to improve the effectiveness of full-text retrieval. II. On the effectiveness of ranking algorithms on full-text retrieval

Abstract

In Part I of this study [2], it was found that full-text retrieval resulted in significantly higher recall and lower precision, compared with that of paragraphs, abstracts, and controlled vocabularies. In the present article (Part II of the study), document-term-weighting algorithms proposed in past research for automatic extractive indexing were examined as a means to improve the low precision of full-text retrieval. As a laboratory experiment, the precision of full-text retrieval enhanced by 29 algorithms was examined at the lowest, middle, and highest levels of recall, and compared with that of full-text retrieval without algorithms and to paragraph retrieval. The over-all effectiveness of seven representative algorithms was displayed and investigated in recall and precision graphs. Twenty-nine algorithms significantly improved the low precision of full-text retrieval, and 22 of 29 algorithms achieved higher precision than that of paragraph searching at the same level of recall, although these results were not statistically significant. However, there was no significant difference between algorithms, although the precision of the 29 algorithms varied considerably. The relative performance of algorithms seemed to depend on the search strategy employed and on the level of recall at which precision was measured. © 1988 John Wiley & Sons, Inc.