The methodology behind the product, in plain English. Helpful if you're comparing technical approaches.
- Data source
- Monitors 1 billion sources including social media (X, Facebook, LinkedIn, Instagram, YouTube, Pinterest, TikTok, Reddit, Medium), review sites (G2, Trustpilot, Capterra, Booking), news outlets, forums, blogs, HackerNews, and GitHub.
- Update frequency
- Real-time alerts delivered within minutes via email, Slack, or mobile app. Historical data accessible back 24 months on Company plan.
- Scope and limits
- Boolean Search supports up to 2,000 characters per query. Competitive monitoring supports up to five competitors. API access available. White-labeling and approval workflows included on higher plans. Reports exportable as links, PDFs, PPTXs, and CSV/XLS. Supports 249 languages for matching, with sentiment in 31 languages.
- Technical approach
- Sentiment analysis uses a 278M-parameter Transformers-based model pre-trained on 2.5TB of data across 100 languages, fine-tuned with 500K mentions, achieving 82% accuracy (F1 score) on the five main languages. Emotion Analysis (October 2024) recognizes seven emotions: joy, sadness, anger, surprise, disgust, fear, neutral. AI Alert Overview generates daily summaries; AI Highlights detects abnormal trend variations.