You can use the Sentiment widget to view any entities with sentiment, across all documents, ranked by score or frequency, or various sentiment statistics. This is useful for viewing high level sentiment of entities at a glance.
You can view sentiment as a Bar chart visualization ranking entities according to positive, negative sentiment, or aggregate sentiment.
To view sentiment
- From the sentiment widget, observe that the bar chart is color coded to display positive, aggregate and negative sentiment.
- A common action is to use the "Rank By" dropdown to rank by positive sentiment, or negative sentiment.
After you have made your rankings, the sentiment widget updates to display the entities with the newly adjusted sentiment factoring.
A common activity for sentiment analysis involves tracking changes of sentiment over time. This time-frame based approach provides more accurate results, and could form the basis for deeper investigation, or case management efforts.
To track changes in sentiment
- From the sentiment widget, filter your results based on keywords of interest
For example, running keyword-based queries against the Enron emails data set reveals that the "environment" was a topic that stirred up a significant amount of negative sentiment at the company
2. Using the advanced query builder, further narrow the query using specific time ranges.
This enables the analyst to hone in on the precise time period when sentiment around the "environment" began to change at the company.