SS

SilentStigma

Study of mental health stigma in public conversations on YouTube

Language lens
Stigma insights
Overview
Impact
Stigma map
Cluster explorer
Methods
Technical view
Explore Experiences Through Language

Discover how people describe shared experiences

This search surfaces anonymized language patterns from public writing. Results reflect shared experiences, not individual advice.

9,382 searches
Try:

Stigma insights

Short sketches of patterns that appear in the clusters.

Fear of burden

Many comments show fear of adding pressure on family or friends.

Self blame

Some comments frame distress as a personal flaw.

Supportive replies

Other threads show simple care and encouragement.

Dismissive tone

Some clusters play down symptoms or make jokes about them.

Coping patterns in this sample

Counts of coping related language across all clusters.

Loading coping patterns...

Overview metrics

-

Total Comments Collected

-

Comments in Analysis Sample

-

Sampling Rate

-

Videos Analyzed

-

Avg Comments per Video

-

Channels

-

Discourse Clusters

-

Noise / Outlier Share

-

Avg Cluster Size

-

Largest Cluster Size

How to read this view

  • Each cluster is a group of comments that share similar language.
  • Counts are for samples not for all comments on YouTube.
  • Values describe themes at group level not single people.

Impact

These counts give a rough sense of how often the data and views are used.

6,503

Dataset downloads

21,042

Exploratory sessions

How this can be used

  • Public health teams can spot common fears or myths that appear in comments.
  • Advocacy groups can see which videos invite support and which draw stigma.
  • Teachers can use clusters and quotes to show how stigma appears in daily talk.

Stigma map

Each point is one comment in a common space. Close points use similar language about mental health.

Loading visualization...

Clusters

Filter
Loading clusters...

Methods

Short view of how the pipeline works without code detail.

  • Text: comments are cleaned and filtered for language and spam.
  • Embeddings: each comment is turned into a vector that captures meaning.
  • Clusters: comments with close vectors are grouped into themes.
  • Map: a two dimensional view makes clusters easier to see.
  • Patterns: keywords and phrases give each cluster a short label.
  • Ethics: no diagnosis and no prediction of risk for any person.

Technical view

Very short view of the software stack for readers who care about code.

  • Data: comments and videos are stored in a SQLite database.
  • Models: the system uses a sentence embedding model and clustering.
  • Server: a Flask app serves JSON for plots and tables.
  • Client: the front end uses Plotly and simple JavaScript for views.
  • Config: one YAML file controls main options and paths.

Ethical Notice

  • Research only: for aggregate social science work.
  • No profiles: no tracking of single people.
  • Public data: only public comments and no clinical use.