Glypse
Social Intelligence

Getting Started

How the Social Intelligence dashboard uses questions and queries to build an analysis of matching posts.

The Social Intelligence dashboard analyzes public English-language Bluesky posts. Its charts summarize the volume, sentiment, emotions, and topics of posts matching your queries and time range. The matching posts appear below the charts, where their text provides context for the measurements.

An analysis starts with a query: the words and phrases that determine which posts to include. Ask a question lets you describe a subject or comparison and have AI write those queries. Write queries lets you enter the terms directly or edit generated ones. The toggle above the input switches between these two modes.

Both modes are available without an account. A free account adds query history, bookmarks, and access to more posts in the feed.

The Social Intelligence dashboard

Three areas make up the dashboard:

  • The query bar at the top contains the question or queries, the time range, and the topic filter.
  • The results area presents summary statistics and charts after an analysis runs.
  • The Recent posts feed below the results lists the individual posts included in the analysis.

A row of tabs appears beneath the query bar once you have results. Each tab examines a different aspect of the same set of posts, from the amount of discussion to the emotions expressed. The charts provide an overview of that discussion, while the feed lets you read the posts behind the figures.

Questions and generated queries

A request such as “Compare sentiment around ChatGPT and Gemini” asks for a separate query for each subject. The generated terms remain visible, so you can check whether they capture the discussion you meant and change the wording if needed.

The time range and topic filter above the input are separate settings. Their Apply buttons save your selections for the next analysis; mentioning a period or topic filter in the question does not set those controls. Every generated query uses the range and filter applied there.

Analyze sends the request, as does Enter from the question field. Once the queries are ready, they appear in Write queries and the analysis runs automatically. A first analysis opens Overview. If you were already reading another results tab, the new results appear in that tab.

A note accompanies the generated queries. It repeats the request under You asked, explains the chosen terms under Why these queries, and describes what the queries match. On a phone, What they match opens any descriptions folded beneath the queries.

The example requests beneath an empty question field demonstrate the kinds of analysis you can ask for. For your own request, Shift+Enter adds a line break without submitting the text.

Revising an analysis or changing the subject

Once you have results, the next request can build on the current queries. Revise in the generated-query note returns to Ask a question, where you can describe a change such as “Include the ticker symbols” or “Add Nvidia to the comparison.” Analyze sends that request together with the current valid queries, including any manual edits. A successful response updates the queries and runs the revised analysis.

The preview beneath the question field shows the queries currently in the bar. New question changes the next request from a revision to a replacement, which is useful when you are moving to a different subject. The preview then says Your next request will replace these queries. The current queries and results stay in place until the new request succeeds. Revise in the preview switches back to making changes to the existing queries.

You can also edit the terms yourself in Write queries. The Edit queries button beneath the preview and the individual preview rows open the same inputs. Switching modes preserves the queries and results. The × on the generated-query note hides its explanation while leaving the analysis in place.

Cancelled or unsuccessful requests

While the queries are being written, Cancel stops the request. Escape does the same when the question field has focus. Your request text and any previous queries and results remain available.

If a request does not name a subject, the field asks for a person, company, or topic. Other unsuccessful requests also leave the text available for editing and resubmission. If you reach the request limit, a message asks you to wait before trying again. Write queries remains available for entering terms directly.

Writing queries directly

Write queries gives you direct control over the words or phrases that posts must match. A query can describe a company, person, product, event, or any other subject discussed in the available posts. A name or a few keywords are enough to begin; operators let you narrow or broaden the matches when needed.

In this mode, Analyze runs the queries in the inputs with the selected time range and topic filter. Enter from a query input does the same.

Basic query syntax

When a query contains more than one word, all of them must appear in the same post. For example, AI energy matches posts containing both “AI” and “energy,” in either order. Matching is case-insensitive. The following symbols change how terms are combined:

MatchSyntaxExample
All termsA space between termsAI energy
Either term, or bothA pipe between alternativesvegetarian | vegan
A phraseStraight double quotes around the words"electric vehicles"
A term while excluding anotherA minus sign immediately before the excluded termcrypto -bitcoin

Quotation marks keep words together in the stated order. Without them, electric vehicles can match a post where those words appear far apart. An exclusion removes the whole post: crypto -bitcoin omits posts containing “bitcoin,” including posts that also discuss other cryptocurrencies.

Operators can be combined. "electric vehicles" -batteries requires the phrase “electric vehicles” and excludes posts containing “batteries.” The Query Syntax reference explains more complex combinations, grouping, and matching hashtags, cashtags, and handles. These rules also apply when you edit a query generated through Ask a question.

Editing and rerunning queries

Editing a query prepares the next analysis. The charts and posts continue to show your last submitted queries until you run the changes. Edited rows are highlighted, and Undo restores a row's last submitted text. A row's hits count describes its submitted results; it disappears while that row's text differs from the query that produced them.

The Clear query control empties one row and its display name. Remove query in the row's three-dot menu removes the row from a comparison. An empty extra row does not add another set of results when the queries run.

If the input reads like a question, the bar explains that its words are being treated as query terms. Ask a question moves the text into the question field for you to review and submit. Keep as a query dismisses that suggestion when you intend to use the text as written.

Time range and topic filters

The calendar control above the input selects the period to analyze. Apply saves the range selected inside the control, while Cancel closes it without applying that selection.

The range determines which posts contribute to every count, score, and distribution. A short range can isolate a particular change; a longer one includes more of the surrounding discussion. If a subject attracted negative reactions yesterday, those posts might dominate the sentiment score for a short range but have less effect on a score that includes several weeks of earlier posts. The time charts show how the discussion changed within the selected period.

The Filter button beside the calendar opens a checklist of topic labels, such as Technology or Business. Apply sets the filter for the next analysis. A post must then have at least one selected label, match the query, and fall within the time range. All and None both remove the topic restriction. The same range and topic filter apply to every query in a comparison.

Applying either control prepares the settings for the next analysis. The results continue to show the previously submitted settings until another analysis runs, and controls that differ from the displayed results are highlighted. In Write queries, Analyze reruns the existing queries with the new settings. In Ask a question, it submits the request in the question field, and a successful response runs the resulting queries with those settings.

Topic filtering selects whole posts and retains all their labels. A post labeled both Technology and Business still contributes both labels when Technology is selected. The Topics reference explains this distinction and the categories used for topic classification.

Comparing subjects

An analysis can contain several queries, each representing a subject or a different way of selecting posts. In Write queries, Add query beneath the inputs creates another row; its label is Add on a phone. The additional query joins the analysis the next time Analyze runs. A request in Ask a question can also produce a comparison. Each query has a color and shape that identify it in the results.

With several queries active, the results offer three views:

ViewWhat it shows
CompareEach query's measurements together in the charts. This is the initial view after adding a query.
AggregateOne combined analysis of posts matching any active query. A post matching several queries is counted once.
SelectThe full results for one query, chosen from the dropdown.

A post can match more than one query in Compare. Adding the separate post counts can therefore exceed the distinct-post total shown in Aggregate. The feed combines posts from all queries in Compare and Aggregate, with shared posts shown once. Select restricts the feed to the chosen query.

For longer queries, the three-dot menu on each row has an Edit display name option. A short name replaces the query text in chart labels and legends. If the row still contains its submitted query, applying a name updates those labels immediately. If you have also edited the query text, the name applies when you run it. Renaming does not change which posts the query matches.

Prepared comparisons

The starting page's Explore by topic section offers comparisons with queries already written for their subjects. Each card previews post counts, sentiment scores, and changes in those scores from the preceding period. Pointing at a subject shows its query; the column headings explain the measurements and the period they cover.

Opening a card runs its comparison and displays the queries in Write queries, where they can be edited in the same way as your own. A card can still open when its preview figures are temporarily unavailable.

Charts and matching posts

Stats bar and charts

The Overview tab brings the main measurements together. For one query, its stats bar shows the matching post count, sentiment score, most common emotion, and leading topic. These summary figures describe the conversation over the selected time range.

The charts below the stats bar show the distributions behind those figures. A sentiment score describes the balance between positive and negative posts; the distribution chart also shows how much of the conversation is neutral. Similarly, the leading emotion names the most common label, while the emotion chart shows the other feelings present in the same discussion. The volume chart shows whether activity was spread across the period or concentrated around particular times.

Sentiment classification describes each post's positive, neutral, or negative tone. Emotion labels describe more specific feelings, and both apply to the post as a whole. A post that mentions two subjects carries the same labels into both subjects' query results. Reading the text helps identify which subject a reaction concerns. The Sentiment reference explains the score and the distinction between a post's tone and an opinion about a particular subject.

Analytical tabs

The tabs provide more detail about each aspect of the conversation. Moving between them keeps the last submitted queries, time range, and topic filter, so switching tabs does not apply pending edits in the query bar.

TabWhat it shows
OverviewSummary statistics, sentiment and label distributions, post volume over time, and a word cloud.
VolumePost counts over time, peak and average volume, and patterns of activity.
SentimentThe sentiment score, the balance of positive, neutral, and negative posts, and changes over time.
EmotionsPost counts by emotion, the diversity of emotion labels, and emotion trends over time.
TopicsPost counts by topic category, the diversity of topic labels, and topic trends over time.
MentionsThe hashtags, cashtags, link domains, and handles found in matching posts, ranked by frequency.

Mentions has a separate toggle for Hashtags, Cashtags, Links, and Handles. Its word cloud, treemap, and ranked table show which entries are most common. Entries can also be added to a query to examine a narrower part of the discussion; the Mentions reference explains those actions and how they work in comparisons.

Recent posts and filters

The Recent posts card below the charts lists matching posts with the newest first. Each row shows the post's text and timestamp, with a tint for sentiment and badges for detected emotions. The Topics tab adds topic badges. These labels help relate individual posts to the distributions shown above them.

The feed's Filter by sentiment or emotion box narrows the posts already loaded by sentiment or emotion. Selecting several sentiments accepts any of them; selecting several emotions accepts any of those. If you select both kinds of label, a post must match a selected sentiment and at least one selected emotion. For example, Negative and Anger show posts with both labels. These filters change the feed alone, so the charts continue to summarize the full query.

When more posts are available, signed-in users can click Show more. Visitors without an account see a sign-up option in its place. A filtered feed can be empty while older posts remain available through Show more. Clear filters in an empty filtered feed removes those label selections.

The feed also helps explain unexpected results. If there are few or no posts, removing an unnecessary term or choosing a longer available range can broaden the analysis. If the posts concern a different meaning of your subject, a phrase or an additional term can narrow the match. In Write queries, you can adjust the terms and use Analyze to update the charts and feed together.

Post details

Clicking a post opens a panel on the right with its full text, timestamp, and available embedded media, such as images or link previews. The author's display name and handle link to their Bluesky profile. Available reply, repost, and like counts appear beneath the post. The panel uses the same sentiment tint and emotion labels as the feed.

The View on Bluesky link opens the original post. This can provide further context, such as replies in the surrounding conversation, when a chart or classification label needs a closer look.

Starting over and clearing inputs

Start over is available in both modes. It returns to the starting page and clears the current analysis, question, and query inputs. It also restores the default time range and resets topic filters, chart controls, and feed selections. Your remembered choice of Ask a question or Write queries remains, and saved history and bookmarks are kept.

The other controls have narrower effects:

ControlEffect
New question in Ask a questionPrepares the next request to replace the queries. The current results, time range, and topic filter remain in place.
Clear all in Write queries, shortened to Clear on a phoneLeaves one blank query row and clears the topic filter. The time range and existing results remain until another analysis runs.
The × on the generated-query noteHides the explanation without changing the queries or results.

Query history and bookmarks

Query history and bookmarks are available with a free account. The history button (a clock icon) above the inputs opens your query history; Ctrl+K or Cmd+K opens it from the keyboard. Its search field matches query text and display names. Selecting an entry restores the queries, names, time range, and topic filter and runs the analysis immediately.

The star button bookmarks the queries and filters used for the displayed results, so it does not save changes that you have yet to run. Bookmarked entries appear above recent queries in the history dialog. Running the same query set with a different range or topic filter updates its history entry to those settings.

Opening a saved query runs it again on the latest posts, so with a rolling time range its counts and posts can differ from last time.

Reference pages

The reference pages explain the query syntax and the measurements in each analytical view, including how their charts change when comparing several subjects.

Further reading

The Sentiment Analysis guide explains how text classification produces sentiment labels and scores, and how volume and the distribution of labels affect their interpretation. The Emotion Classification guide covers the specific feelings represented by emotion labels and why a post can express more than one at a time.