| Home | Organizers | Proceedings Editors | Proceedings Contributors | Search | | ||
|
|
|||
|
||||
|
|
||||
| Title: | AUTOMATIC UNSUPERVISED KEYPHRASE-BASED QUERY EXPANSION FOR BIOMEDICAL DOMAIN | |
| DOI No: | 10.1142/9789812701527_0018 | |
| Source: | KNOWLEDGE MANAGEMENT: NURTURING CULTURE, INNOVATION AND TECHNOLOGY (pp 209-220) | |
| Author(s): | MIN SONG
College of Information Science and Technology, Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104-2875, USA IL-YEOL SONG College of Information Science and Technology, Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104-2875, USA KI JUNG LEE College of Information Science and Technology, Drexel University, 3141 Chestnut Street, Philadelphia, PA 19104-2875, USA |
|
| Abstract: | This paper introduces an automatic querying technique, DocSpotter, and presents that it is an efficient tool to identify the set of documents for the extraction of a pre-defined relation from text. DocSpotter is designed to retrieve documents with more precise match to the initial query by expanding queries as it repeats the keyphrase extraction process in a given database. It utilizes keyphrase extraction in conjunction with referencing ontologies for the query expansion. We report two sets of experimental results demonstrating the performance of DocSpotter. The experiments were designed to evaluate the performance of DocSpotter on the task of protein-protein interaction extraction. The results identified that DocSpotter was able to retrieve more and more documents that contain protein-protein pairs from MEDLINE as it repeated the keyphrase extraction process. In the other set of experiments, performance of DocSpotter was compared with that of SLIPPER, a supervised rule-based query expansion technique. The results showed that DocSpotter outperformed SLIPPER from 17.90% to 29.98% in terms of accuracy in all iterations. | |
| Full Text: | View full text in PDF format (814KB) | |
| TOC: | Back to Table of Contents | |
|
||