Yunli Wang1, Zhenkai Liu. 1. Institute for Information Technology, National Research Council Canada, 46 Dineen Drive, Fredericton, NB, E3B 9W4 Canada. Yunli.wang@nrc-cnrc.gc.ca
Abstract
OBJECTIVES: To automatically evaluate the quality of health information on the Internet, we presents a method for detecting indicators for quality of health information. METHODS: An automatic indicator detection tool (AIDT) was developed in the following steps: (1) 18 initial technical criteria were chosen; (2) multiple measurable indicators were defined for each criterion; (3) these measurable indicators on the Web pages were automatically detected by a computer program; (4) the detectable indicators were selected for each criterion based on detection accuracy and occurrence frequency, and detecting criteria on Web pages; (5) AIDT was developed based on the detectable indicators and criteria. The efficiency of detection tool was measured by recall and precision of detecting criteria. The performance of AIDT was tested using three data sets. RESULTS: AIDT with validated detectable indicators can reach 93% recall and 98% precision on these data sets. AIDT can automatically and correctly detect measurable criteria on Web pages. CONCLUSIONS: Automatic detecting indictors for quality of health information is feasible and effective. Using an automatic detecting tool is a promising method for identifying quality criteria of health information, and eventually it can be used to develop instruments for evaluating the quality of health Websites.
OBJECTIVES: To automatically evaluate the quality of health information on the Internet, we presents a method for detecting indicators for quality of health information. METHODS: An automatic indicator detection tool (AIDT) was developed in the following steps: (1) 18 initial technical criteria were chosen; (2) multiple measurable indicators were defined for each criterion; (3) these measurable indicators on the Web pages were automatically detected by a computer program; (4) the detectable indicators were selected for each criterion based on detection accuracy and occurrence frequency, and detecting criteria on Web pages; (5) AIDT was developed based on the detectable indicators and criteria. The efficiency of detection tool was measured by recall and precision of detecting criteria. The performance of AIDT was tested using three data sets. RESULTS: AIDT with validated detectable indicators can reach 93% recall and 98% precision on these data sets. AIDT can automatically and correctly detect measurable criteria on Web pages. CONCLUSIONS: Automatic detecting indictors for quality of health information is feasible and effective. Using an automatic detecting tool is a promising method for identifying quality criteria of health information, and eventually it can be used to develop instruments for evaluating the quality of health Websites.