Sentiment
The Sentiment tab summarizes the positive, neutral, and negative labels assigned to matching posts.
The Sentiment tab of the Social Intelligence dashboard shows the balance of positive and negative posts, the share labeled neutral, and how those counts change over time. Its results use the submitted query, time range, and topic filter. Getting Started explains these controls, and Query Syntax explains how to select the posts to analyze.
Overview summarizes sentiment across the selected period. This tab adds the sequence behind that summary. A negative full-range score can come from consistently negative intervals, a recent decline, or an earlier drop followed by an improvement. The time-series chart shows which pattern the matching posts follow.
Each post receives one sentiment label for its text as a whole. The same label contributes to every query that matches that post. A post can discuss several subjects, so reading the text helps identify which subject an evaluation concerns. The Sentiment Analysis guide explains the distinction between whole-post and target-specific classification.
Reading the score and distribution
With one query, the tab shows a stats bar, a Sentiment score gauge, a Sentiment distribution pie, and a time-series chart. Select reproduces this layout for one chosen query when several are active.
The score runs from -100 to +100. It subtracts the negative post count from the positive count, divides by their combined count, and multiplies by 100. Neutral posts are excluded from this calculation.
For example, a result with 60 positive, 20 negative, and 120 neutral posts has a score of +50. The calculation uses the 80 positive or negative posts. The pie chart shows the full distribution: 30% positive, 10% negative, and 60% neutral. Hovering over a pie segment shows its post count and percentage.
A score of +100 means there are positive posts and no negative posts. A score of -100 means there are negative posts and no positive posts. Zero means the positive and negative counts are equal, or that both counts are zero. An all-neutral result therefore displays zero too; the pie chart distinguishes these cases.
The Sentiment stat above the gauge gives the score a level label. The same levels determine the gauge's color.
| Level | Score range |
|---|---|
| Very Positive | Above +50 |
| Mostly Positive | Above +25, up to and including +50 |
| Neutral | -25 to +25, inclusive |
| Mostly Negative | -50 up to, but excluding, -25 |
| Very Negative | Below -50 |
These five levels describe the aggregate score. Individual posts have the three labels positive, neutral, and negative. A Neutral level can summarize a mixture containing many positive and negative posts; it does not mean that every post was labeled neutral. Levels use the unrounded score, while the gauge and stat display a rounded whole number.
The score preserves the ratio between positive and negative counts. In the example, 600 positive and 200 negative posts would also produce +50. The smaller result changes more when one additional positive or negative post arrives. The score alone gives neither the number of contributing posts nor the intensity of their language, so the distribution and counts are part of reading it.
Following sentiment over time
Sentiment over time groups posts into days or weeks, with the interval chosen automatically from the selected range. Its three modes show different measurements of the same intervals.
| Mode | What it shows |
|---|---|
| Score | A line of interval scores using the same positive-negative calculation as the gauge. This is the initial view. |
| Volume | Stacked counts of positive, neutral, and negative posts in each interval. |
| Distribution | Each category's percentage of the posts in that interval, adding to 100% when the interval contains posts. |
The Score line puts the highest and lowest intervals in context. A low point may be isolated, or it may belong to a run of falling scores. Crossing zero means the balance between positive and negative post counts changed sides; the Volume view shows how many posts underlie that change.
Volume explains which counts changed beneath the score. Suppose one day contains 60 positive and 20 negative posts, followed by a day with 20 positive and 20 negative posts. The score falls from +50 to zero because positive volume fell. Negative volume stayed constant. Distribution shows the changing proportions, while Volume preserves that difference in counts.
An interval with no positive or negative posts receives a score of zero. That can mean an all-neutral interval or an interval with no posts at all. Volume distinguishes both from an equal positive-negative split. Scores based on a few positive or negative posts can also move substantially after a small change in counts.
The full-range score combines post counts across the range. Intervals containing more positive and negative posts contribute more to that calculation. A simple average of the points on the line would give every interval equal weight. If the final interval is still underway, its chart values describe the part available so far.
Reading the stats bar
The Sentiment stat reports the full-range score and its level. The remaining cells summarize the time series. The highest, lowest, and trend calculations leave out the final interval when there is more than one. The charts and full-range score retain that interval, which can explain a difference between the last point on the chart and the summary cells.
Most positive day or week identifies the interval with the highest score. Most negative day or week identifies the lowest. Each shows a date and a level label. These are relative rankings: if every interval has a negative score, the most positive one has the least negative score. Hovering over the time-series chart shows the score for each interval.
Sentiment trend is the average change in score between successive intervals, measured in points per day or week. Positive values are labeled Improving, negative values Declining, and zero Stable. Suppose the intervals included in the trend calculation have scores of +10, +30, and +20. Their changes are +20 and -10, averaging +5 points per interval. A positive trend can therefore contain declines along the way. Stable can also accompany rises and falls that cancel out, or too few intervals to calculate a trend.
The full-range level and the trend describe different aspects of the result. A Mostly Negative score can accompany an Improving trend, and a Mostly Positive score can accompany a Declining trend. The first summarizes the collected posts; the second describes the average direction of change between intervals. The dates of the highest and lowest scores provide starting points for inspecting the posts behind that movement.
Comparing queries
With two or more submitted queries, Compare displays their scores and distributions together. Every query uses the same time range and topic filter.
Comparison stats
Most positive and Most negative identify the queries with the highest and lowest scores. The rankings are relative, so Most positive can name a query whose score is below zero. Queries with very different post counts can share a score because they have the same positive-negative ratio.
Most neutral and Least neutral identify the largest and smallest neutral shares. Each share divides neutral posts by all positive, neutral, and negative posts for that query. It describes the label mix, without measuring how much the authors care about the subject or how strongly they hold their views.
Comparison charts
The Sentiment chart opens in Score mode, with one bar per query on the -100 to +100 scale. Distribution divides each query's bar into positive, neutral, and negative percentages. Switching between them shows how much of each result contributes to its score. Each query's Select view provides the underlying counts in its distribution pie; hovering over a segment shows the count and percentage.
Sentiment over time draws one score line per query. This comparison chart has no Volume or Distribution mode; those modes are available in each query's Select view. A crossover shows a change in the relative scores. To investigate it, Select opens each query's full charts and matching posts.
The results can overlap. A post matching two queries contributes the same sentiment label to both. Similar lines may therefore partly reflect shared posts. The charts alone do not establish that the subjects received the same reaction or that a shared event caused their scores to move together.
Combining queries with Aggregate
Aggregate combines posts matching any active query, counts each matching post once, and shows the single-query layout. It recalculates the score from the combined positive and negative counts. Averaging the separate query scores would give a different calculation, especially when their counts differ or their results overlap.
This view can combine alternative names or spellings for a subject. The score still describes the whole-post labels of the combined result, so reading examples helps establish which discussions the added queries brought in.
Inspecting the posts behind a change
The query bar's calendar control can narrow the analysis to a particular interval. After you apply a range, Analyze in Write queries updates the results for that period. With several queries, Select limits the charts and feed to the chosen query. The Recent posts feed below the charts contains the text behind the measurements; clicking a post opens its details.
The feed's Filter by sentiment or emotion box can show posts labeled Positive, Neutral, or Negative, and can combine sentiment with emotion filters. It filters posts already loaded and leaves the charts unchanged. Getting Started explains the filter combinations and how to load more posts. The Emotions tab provides a further breakdown of the feelings expressed in the same collection.