Figure 2. Exploring journal entries using text sentiment and keyword extraction.
As shown in Figure 2, an interface was developed to allow exploration of journal texts by sentiment and
keywords. Using Pattern, a python library for Natural Language Processing, we extracted sentiment for
each dated journal entry and plotted the score against time1. A LOESS curve is then drawn (within the
browser app using science.js) which indicates a steady, positive sentiment associated with using mobile
devices in the classroom.
1 Pattern’s sentiment module uses a lexicon of adjectives that are annotated with scores for sentiment polarity positive to negative). The module
returns a score between - 1.0 and + 1.0, which is the average of scores across all adjectives used in the text.