Frequently Asked Questions (FAQ)

Does PJQ store a database of user comments?

No. PJQ processes comments in RAM during the classifier's operation. Only the final mathematical percentages and indexes, along with 2-3 anonymized text examples per category for display in the Seismograph, are saved to the database. We do not build dossiers on users.

How accurate is the classification?

The PJQ model is trained on complex semantics and outperforms traditional sentiment dictionaries because it analyzes text while accounting for context (video title and domain). The algorithm can distinguish genuine positivity from sarcasm (the AGA category) and detect covert self-promotion (the AGN category).

How is the Featured showcase on the homepage generated?

The video showcase on the main page is generated strictly algorithmically based on a final ranking formula. PJQ has no manual curation or promotion of specific videos—this guarantees the platform's neutrality.

How does this differ from YouTube Analytics?

Built-in analytics show engagement metrics: how many people clicked, how many left a comment. PJQ shows the quality of that engagement: how many are discussing the topic (NEU), who came to support it (SUP), and how many are just spamming with fake links (SUS).

How does this differ from classic sentiment analysis?

Classic sentiment analysis sorts comments by tone: positive, negative, neutral. PJQ measures stance toward the author's thesis: substantive agreement and disagreement are separated from fan praise, spam and self-promotion, so the scores reflect opinion rather than noise. And the workflow stays simple — "URL in → Verdict out", with maximum respect for privacy.