Regresar

Mining twitter for measuring social perception towards diabetes and obesity in central america

Abstract:

For a long time, diabetes and obesity have been considered a menace only in developed countries. Nevertheless, the proliferation of unhealthy habits, such as fast-food chains and sedentary lifestyles, have caused diabetes and obesity to spread worldwide causing many and costly complications. Since citizens use of the Internet to search, learn, and share their daily personal experiences, the social networks have become popular data-sources that facilitate a deeper understanding of public health concerns. However, the exploitation of this data requires labelled resources and examples; however, as far as our knowledge, these resources do not exist in Spanish. Consequently, (1) we compile a balanced multi-class corpus with tweets regarding diabetes and obesity written in Spanish in Central-America; and, (2) we use the aforementioned corpus to train and test a machine-learning classifier capable of determining whether the texts related to diabetes or obesity are positive, negative, or neutral. The experimental results show that the best result was obtained through the Bag of Words model with an accuracy of 84.30% with the LIBLinear library. As a final contribution, the compiled corpus is released.