What does Zencity mean by sentiment?
Our sentiment analysis algorithm determines if data reflects positive or negative feedback about local government-related issues. Every conversation in the platform, be it a post or a comment, is assigned a sentiment score, based upon the content of its text and accompanying context of the likes, comments, and shares it receives. If residents are expressing satisfaction with the issue in question, then the sentiment will be marked as positive; if residents are dissatisfied, then it will be negative.
There is no difference if negative feedback stems from the issue being discussed itself, or from the city’s role in the issue - the content will carry negative sentiment regardless. The reason for this is that it is extremely hard for an algorithm to differentiate between the two scenarios, yet we want to ensure concerns that are top-of-mind for residents are flagged. By contrast, we consider a post or a comment that contains purely informative data as carrying no sentiment and we'll most likely classify it as neutral.
How does Zencity’s AI assign sentiment to a text?
Sentiment is not assigned based on the tone of the words alone. Zencity’s AI is conversation-based and takes into account the context of a piece of text in its entirety when assigning sentiment. We’ve learned that what makes content positive or negative for cities isn’t necessarily the nature of the words or events.
For example, while a break-in carries negative connotations, a post informing the public about a break-in that occurred shouldn’t be classified as negative (unless explicit dissatisfaction is expressed in the text). And in fact, a conversation where people express gratitude about a prevented break-in might be considered positive feedback.
Our sentiment analysis is driven by AI and constantly trained by experts. While individual mistakes are possible, the algorithm should pick up on the general sentiment trend of most conversations, and is always improving over time. If you notice items that seem to be incorrectly tagged, you can always let your Customer Success Manager know or update the sentiment label yourself.
What does “neutral” mean and why is it interesting?
Neutral is discourse without sentiment, as mentioned above. Whereas positive or negative sentiment can be used to pinpoint specific issues, concerns, satisfaction, or feedback related to posts, neutral represents discourse that does not specifically relate to a sentiment.
If your post is trending but it’s mostly neutral, it means that it received a lot of attention but not necessarily expressions of positive or negative feedback. In some cases, this can be a helpful indicator in measuring the effectiveness of a social media campaign, in which a high volume of neutral sentiment may represent success.
How can I use sentiment to easily find the issues that are concerning to residents?
First of all, events that cause a strong, negative public reaction are easy to spot via negative sentiment tracking. For example, an instance of police misconduct may generate negative sentiment in the police category. To watch for spikes in negative sentiment or to pinpoint specific resident concerns, you can:
Scan the updates in Today's Highlights
Take a look at the red line in the volume of interactions graph
Filter by negative sentiment in the main dashboard or within a project
How can I use sentiment to assess resident feedback on an issue or policy?
Creating a project in Zencity will consolidate discourse on a specific issue, event, or policy. Within the project, you can then filter by positive, negative, and neutral sentiment to gain visibility into all angles of the conversation.
How can I use sentiment to see how residents react to posts I publish?
First use the Channel filter at the top of your main or project dashboard to display only content aggregated from official channels (i.e. content from any pages or accounts you manage). Find the specific post you're interested in following under the Stories tab, click on it to open up the full thread, and then filter comments by positive, negative, or neutral sentiment.
How can I use sentiment to get a sense of what residents care about?
The most popular topics widget within your main dashboard ranks the topics are residents are talking about according to engagement (volume of interactions). Next to each topic is a breakdown of the correlated positive and negative resident, so you can immediately get a sense for how residents feel about some of the most pressing current issues.