InDeCS: Método automatizado de classificação de páginas Web de Saúde usando mineração de texto e Descritores em Ciências da Saúde (DeCS) / InDeCS: Automated method for classification of health Web pages using text mining and Health Sciences Descriptors (DeCS)
Falcão, Alex Esteves Jaccoud; Mancini, Felipe; Costa, Thiago Martini da; Hummel, Anderson Diniz; Teixeira, Fabio Oliveira; Sigulem, Daniel; Pisa, Ivan Torres.
J. health inform
; 1(1): 1-6, 2009.
Article
in Portuguese
| Redbvs | ID: biblio-859074
The amount of webpages has growing strongly, potentially leading knowledge to more people, but with the disadvantage of hindering relevant and reliable information.
Objective:
To present results of an automated method to classify and indexing health webpages.Methods:
It was selected and classified webpages manually as health (saúde) and non-health (não-saúde). On a second step it was calculated the similarity between the webpages terms and the Health Science Descriptors (DECS). Automated classifiers parameters were developed using these similarities values.Results:
For this experiment were collected 1,132 webpages, separate in "saúde", "não-saúde" and "Merck" databases, generating more than 3 million of 3 grams compositions. The experiment using the "saúde" and "não-saúde" databases resulted hit, sensitivity, specificity and area under ROC curve, respectively, 85.10%, 0.81, 0.88 and 0.92. The other experiment using the "Merck" and "não-saúde" databases resulted respectively, 97.44%, 0.92, 1.00 and 0.98.Conclusion:
The preliminary results of this text mining metric using controlled vocabularies to improve the result of web search engines specifically for health were significant...(AU)
Responsible library:
BR1.1
Fulltext
Related documents