News
A client recently asked me to analyse several thousand website comments to reveal the underlying sentiment.
The challenge was not just the sheer size of the database but also in the vary wide variety of topics that people commented on - usually very briefly.
The analysis process was an iterative one, where I kept going back and forward testing out my emerging ideas on the data. I had to develop strategies for dealing with the size of the database, and for knowing what to ignore - knowing what to ignore is often the most important decision an analyst makes. I call this 'pruning' in my analysis model.
In the end, the underlying sentiment was a shared metaphor.
