What are effective ways to help people with chronic illness, e.g. diabetes and heart disease? Computational linguistics relies on human-annotated data to train machine learners. Inconsistency among the human annotators must be carefully managed (otherwise, the annotations are useless in computation). How can this annotation process be made scalable? ABSTRACT Curing chronic illnesses and diseases requires the huge effort of collecting all available information on this matter and piecing it together with the aids of mathematical and computer modeling. Both phases of information collection and piecing together are prone to error. Errors may result from human annotation inconsistency, machine learning and parameterization when using supervised learning. On a different scale, published results that need to be collected may suffer from another kind of disagreement either due to varying experimental methodologies or assumptions. Here, we discuss these inconsistencies and disagreements in scientific literature and we investigate those of the inter-annotation of named entities in bioliterature from empirical perspectives.