Sentiment, Scoring & Methodology
What does KnowEntry mean by “sentiment”?
At KnowEntry, sentiment refers to the measurable tone of news coverage about AI companies and sub-sectors — not investor positioning or market prices. Canary scores each article on a scale from -1.0 (clearly negative coverage) to +1.0 (clearly positive coverage), with 0 indicating neutral or balanced reporting. The daily sentiment figure for each category is a relevance-weighted average across all articles scored that morning, where relevance is rated from 4 to 10 based on how directly the article addresses the category in question. This is editorial sentiment — how journalists and analysts are framing AI developments — not a market indicator.
How is KnowEntry sentiment different from market sentiment?
Market sentiment reflects investor behaviour, positioning, and prices. KnowEntry sentiment reflects how journalists and analysts frame AI-sector developments in financial and technology news. The two often diverge significantly: news narratives, investor positioning, and prices update on different timelines and respond to different information. A period of strongly positive Canary scores does not imply rising prices, and a period of negative scores does not imply falling ones. Canary measures the narrative environment, not the investment environment.
Is Canary a trading or investment signal?
No. Canary is an interpretive intelligence system, not a trading or investment recommendation tool. It measures narrative tone and narrative change across AI sector news coverage. It does not predict prices, valuations, or market outcomes. Any use of Canary data in financial decision-making is entirely at the user’s own discretion and risk, and no such use is implied or encouraged by KnowEntry.
What are Canary’s eight narrative frames?
Canary assigns every article to one of eight mutually exclusive narrative frames: Growth Momentum, Technical Breakthrough, Financial Results, Regulatory Risk, Geopolitical Risk, Competitive Threat, Market Correction, and Macro Environment. The frame captures what kind of story is being told — not just whether coverage is positive or negative. Two days with identical sentiment scores can represent entirely different situations if the mix of story types differs. Regulatory Risk stories carry different implications than Growth Momentum stories even when both produce a moderately positive sentiment reading.
What is the Semantic Volatility Index (SVI)?
The SVI is a composite early-warning indicator developed for Canary 2.0, scored from 0 to 100. It aggregates five sub-components: Sentiment Volatility (SC1, lagging — rolling standard deviation over a 7-day window), Within-Day Dispersion (SC2, concurrent), Narrative Shift (SC3, concurrent), Vocabulary Drift measured using Jensen-Shannon Divergence (SC4, leading), and Frame Distribution Shift measured using Jensen-Shannon Divergence (SC5, leading). An SVI above 75 indicates extreme volatility and a possible narrative regime shift. The SVI is specifically designed to detect structural changes in the news narrative, often before those changes appear in headline sentiment scores.
What is a narrative regime shift?
A narrative regime shift occurs when the fundamental character of AI sector coverage changes — not just the level of sentiment, but the vocabulary and framing. For example, a shift from earnings-driven Growth Momentum coverage to Geopolitical Risk coverage can occur at similar sentiment levels but represents a qualitatively different environment with different implications. The SVI’s leading sub-components (SC4 and SC5, both using Jensen-Shannon Divergence) are specifically designed to detect this kind of shift early, before it becomes visible in aggregate sentiment readings.
How often is Canary updated?
Canary publishes new scores each morning, based on the previous day’s coverage. The KnowEntry news aggregator sections are updated continuously every 10–60 minutes around the clock. The Canary Dashboard reflects the most recent morning’s scoring run. Historical scores are retained and available for trend analysis.
What six categories does Canary cover?
Canary scores news across six AI industry categories: AI Chips and Hardware, AI Platform Hyperscalers, AI Enterprise Software, Data Centre Infrastructure, AI GPU Cloud, and Pure Play AI. These represent distinct layers of the AI value chain from infrastructure to application. Category-level divergence — where one layer strengthens while others weaken — is often more informative than the overall headline score, since different parts of the stack respond to different news drivers.
What does the 95% confidence interval on Canary scores mean?
Each daily sentiment score is shown with a 95% confidence interval representing statistical uncertainty around that day’s estimate. A narrow interval means the articles scored that day were broadly consistent in their tone. A wide interval means fewer articles were available, or that sources disagreed significantly in their framing. A wide interval indicates uncertainty in the estimate — it is a signal to interpret cautiously, not an error in the system.
How should historical Canary scores be compared?
Canary scores are most meaningful when compared within the context of each category’s own recent history, rather than against other categories or fixed thresholds. Canary uses Welch’s t-test to determine whether any day’s reading is statistically unusual relative to that category’s recent baseline — a p-value below 0.05 means there is less than a 5% probability the reading is random variation. A score of +0.3 may be unremarkable for one category and statistically significant for another. Sentiment direction, narrative frame distribution, and confidence and stability signals should be read together for the most reliable interpretation.
Where does Canary’s data come from?
Canary sources AI sector news from four commercial news APIs — NewsAPI, NewsData, Finlight, and Finnhub — pulling several hundred to over a thousand articles each morning. Articles are deduplicated across sources and dates before scoring. Each article is then assessed for relevance to the AI sector on a scale of 4 to 10, and only articles meeting a minimum relevance threshold are included in sentiment calculations. Typically 500–700 articles are scored each day, from a raw fetch of 800–1,000.