Data
The training set used for the identification of key epidemiological characteristics included the 60 epidemiological abstracts related to obesity from PubMed. Since epidemiological articles are relatively longer than abstracts of other scientific context, manual annotation by the author required a significant amount of time. Hence, the number of 60 abstracts was selected due to the time constraints inflicted upon the completion of this thesis, as an appropriate balance between time investment and sample size. A further set of 30 abstracts (also manually curated by the author) were utilized as a development set to optimize the performance of the proposed methodology. The corpus used for the evaluation of the proposed approach included 60 epidemiological abstracts related to obesity, manually annotated by the author.
