
120 Text Mining and Visualization: Case Studies Using Open-Source Tools
FIGURE 5.20: String length of Description: column.
This provides a good target label to build a prediction on. The Description: field was
chosen as the text field to generate input features for the prediction model. The Description:
free text field was chosen over User Report: as it is mandatory and the latter optional. In
addition it was believed that a label Warning:=Yes may preclude an entry into User Report:
field. (Would a person consume a pill if they believed it to be adulterated?)
The distribution between the labels is No (N=3,184, 64%) and Yes (N=1,817, 36%).
There is no