In this study, the open-source Mallet software package from the University of Massachusetts was applied.
For each document, we removed stop words using a stop word list available at Mallet software package [ 13].
The maximum-entropy (ME) model and Naïve Bayes (NB) model in Mallet toolkit software package [ 28] were applied to construct the classifiers.
Medication entity detection, including medication name and attribute detection, was designed in-house based on the MALLET conditional random field (CRF) package [ 35, 36].
The modified Mallet classification ("Mallet score") (Mallet 1972, Waters 1997) was used as the main outcome measure.
As can be seen in Additional file 5, teams also adapted a range of packages that implement machine learning algorithms (e.g. Mallet or CRF++) or general natural language processing software (e.g. OpenNLP) to build their systems.
They are all checked again before being neatly packaged up to be sent on their way to the distribution centre in Shepton Mallet, also in Somerset, from where they will be shipped around the world.
