Figure 1. Adjective-noun (left) and verb-noun (right) co-occurrence networks of 37 reflective journals maintained by
participants of the Mobile Device Prototype at ASB over nine months.
Verb-noun collocations from the reflective journals may be considered to represent important
information networks for "what we do with what we have". Similarly, adjective-noun collocations may be
considered to represent "how we feel about what we have". We split the raw text into sentences, then
use Patterns Part of-Speech (POS) tagger to extract all verbs/adjectives within a two word gap from every
noun in each sentence. Nouns and verb/adjective collocates form nodes, with links drawn between them
that are sized by the inverse frequency of usage (more commonly used collocates appear closer together).
What emerges in both networks is a single, predominating network of noun-verb and noun-adjective use,
surrounded by smaller node-link pairs at the periphery. This almost immediately allows one to see the 37
individual journals as a single conversation - the main body of the conversation is the most connected part
of the graph, while smaller threads of conversation occur at its edges.
Figure 1 above shows both networks. In the adjective-verb network on the left, participants most
commonly referred to great-app, mobile-app, good-app, different-app, new-app, amazing-app and
educational-app. It is interesting that time (next-time, real-time, huge-time, little-time) emerges as an
In the verb-noun network, it is clear that the journals largely refer to USING, SEEING, HAVING,
RECORDING, TAKING, GETTING, CAPTURING, WORKING, CREATING and MAKING. Of interest is the fact
that meaning circulates differently when participants refer to creating and making - participants tended
to make-sense or make-contact or make-teach while they create-video or create-content or create-groups
or create-quiz/project when using mobile devices in the classroom.
We also created interfaces to explore the data using metrics from text sentiment analysis. Sentiment
analysis (also known as opinion mining) aims to classify the polarity of a text document as positive,
negative or neutral. Because each journal entry is dated, we can explore how sentiment changes over
time. If the prototype was successful, we posit that sentiment, as expressed by a line of best fit across
journal documents through time should have increased or remained positive at the very least.