The Sentiment Analysis widget reveals the most positive and negative tweets. Up to 50 of the most positive and negative tweets are visualized in the graph and listed in their respective columns.
The Pie Chart in the top left part of the graph reveals the overall sentiment breakdown for the selected time period for all tweets. Hover over the two pie slices to get the percentages.
The Bubbles in the graph represents the most significant individual tweets from the time period. The further up on the X-axis the more positive the tweets, the further down on the X-axis the more negative. The size of the bubble indicates the level of engagement – the larger the bubble the more retweets. Select a bubble and the corresponding tweet will be highlighted in the respective columns below.
Sentiment Analysis is powered by a natural language processing (NLP) algorithm optimized for healthcare and is proprietary to Symplur. This algorithm extracts information from healthcare conversations in order to determine polarity about healthcare issues. It takes into account grammar analysis, sentence structure, parts of speech, punctuation, emoticons, slang terms, and shortened terms common in social media.
Each sentiment score is also weighed accordingly based on the tweet author's influence in healthcare.
The method used for determining sentiment employs a scaling system for three classes of neutral, positive and negative sentiment.
- Filter the Healthcare Stakeholders voices by toggling each stakeholder in the toolbar. This filters both the bubbles in the graph and the columns. It will also update the Pie Chart with the average sentiment breakdown of the remaining filtered tweets.
- Select and Drag in the bubble graph to zoom in on a more narrow time period.
- Click a Bubble to see that tweet in the respective columns below.
- Toggle the Lock Switch or Click a Column to enable/disable scrolling inside a column.
- Click the Tweet Timestamp to open the specific tweet on Twitter.
- Click Twitter Handle link to reveal full profile and options to see this user's own tweets, received mentions and its Healthcare Social Graph page.
The sentiment algorithm is highly trainable. The advantage of training the algorithm to your particular context or specialty is that it results in even more accurate sentiment scoring. Training involves either (1) modifying the score attributed to an existing term, and/or (2) adding your own custom term(s) with your assigned score.
To adjust the score of an existing term follow the the steps below:
- Click on the Edit link of the term you want to modify in the table.
- Enter the new score using the input field or adjust the slider to a new score.
- Click on Save when finished. The system will automatically rerun the algorithm with your newly assigned score for the term.
To revert back to the default score of the term, click on the "trash" icon and the system will automatically reset and rerun the algorithm. Note that terms that are in the base dictionary of the sentiment algorithm cannot be deleted, only their scores can be modified.
To add a new term to used by the sentiment algorithm follow the steps below:
- Highlight the word(s) that you want to add.
- Confirm your text selection by clicking on the button that appears below.
- The new selected term will be added to the table where you can assign the corresponding score.
- Click on Save when finished. The system will automatically rerun the algorithm using the newly added term and score.
To delete the custom term click on the "trash" icon and it will be removed from the table and the sentiment score will be recalculated. Note that because this is a custom term, and not part of the base dictionary, it will be deleted.
Sentiment training is only available to Enterprise level clients.