Reference
Glossary
Key terms for deterministic SEC filing analytics and disclosure risk scoring.
- Disclosure risk score
- overall_disclosure_risk_score (0–100) aggregates nine headline-weighted component scores from deterministic_scoring_v2. Higher values indicate more disclosure risk or deterioration on that scale. Not investment advice.
- Item 1A
- Risk Factors section in 10-K and 10-Q filings. Primary source for risk-factor tone, boilerplate, and legal/regulatory language component scores. Validation corpus used Item 1A text.
- MD&A
- Management Discussion and Analysis (Item 7 in 10-K, Item 2 in 10-Q). Parsed for mdna_uncertainty_score and liquidity_stress_score from uncertainty, demand stress, and forward-looking language patterns.
- EDGAR
- SEC Electronic Data Gathering, Analysis, and Retrieval system. Ticker-based scoring fetches filing HTML from EDGAR and requires a valid SEC User-Agent header.
- 10-K
- Annual report SEC form. Supported via ticker routes (CLI, SDK, HTTP, MCP Analyst) or local --html. Primary form for year-over-year disclosure change scoring when a prior filing is supplied.
- 10-Q
- Quarterly report SEC form. Supported via the same surfaces as 10-K. Section taxonomy differs slightly from annual filings; scores include form_type in responses.
- 8-K
- Current report for material events. Scoring supported via local --html or MCP Builder only — not EDGAR ticker routes or HTTP API ticker endpoints.
- Boilerplate risk
- boilerplate_risk_score measures vague, templated risk-factor language in Item 1A, driven by boilerplate_combined_ratio in text_metrics_v4. Validated with Spearman ρ ≈ 0.92 vs ls_boilerplate_word_ratio on boilerplate_combined_ratio (n=478); phrase-only v3 was ≈0.74. Not a full replication of the Lang & Stice-Lawrence paper measure.
- boilerplate_combined_ratio
- Blended boilerplate input in text_metrics_v4 (phrase list + cross-firm 4-gram share) used to compute boilerplate_risk_score. Construct validity checked vs ls_boilerplate_word_ratio.
- boilerplate_phrase_ratio
- Section-level phrase hit rate from a fixed boilerplate phrase list (Lang & Stice-Lawrence-inspired). Phrase-only metrics in v3 correlated ≈0.74 with the LS-style proxy; v4 combines phrase and cross-firm signals.
- boilerplate_cross_firm_ratio
- Cross-firm 4-gram word share in text_metrics_v4 — closer to the Lang & Stice-Lawrence validation reference than phrase-only measures.
- Quintile
- Score distribution bucket (Q1–Q5) used in validation reporting. Q5/Q1 post-filing volatility ratio ≈ 1.15 on n=435 — descriptive association only, not return prediction.
- score_coverage_ratio
- Fraction of expected sections successfully extracted and scored. Low coverage indicates missing sections or extraction issues; inspect section text and extraction_warnings before scaling.
- deterministic_scoring_v2
- Default scoring model version string. Defines component weights and aggregation. Identical input HTML and version strings yield identical scores across all integration surfaces.
- MCP
- Model Context Protocol integration for agent tools. MCP Analyst handles ticker workflows; MCP Builder exposes raw HTML tools including 8-K scoring.
- SEC User-Agent
- Required HTTP header (name and contact email) for EDGAR requests when scoring by ticker. Set via SEC_USER_AGENT environment variable. Not required for local --html scoring.
- section_extractor_v1
- Parser artifact version string in score responses. Extracts stable section IDs (e.g. item_1a_risk_factors) from filing HTML for downstream metrics and scoring.
- text_metrics_v4
- Metrics artifact version string. Computes tone ratios, boilerplate proxies (phrase + cross-firm blend), NER-based specificity, and other text measures on extracted sections.
- built_in_dictionaries_v3
- Built-in word list artifact version. Disclosure Alpha ships built-in lists — not Loughran–McDonald licensed lists unless you load them separately.
- disclosure_change_score
- Component measuring year-over-year section text changes. Requires a prior comparable filing (--prior-html or prior fiscal year). Part of deterministic_scoring_v2 headline weights.
- specificity_quality_score
- Computed component reflecting company-specific vs generic language (higher = more specific). Returned in responses but excluded from headline weight aggregation.
- Panel screener
- POST /v1/panel/disclosure-matrix batch endpoint scoring up to 25 tickers per request. Errors collected per ticker without fail-fast.
- extraction_warnings
- Parser warnings in score responses when section boundaries are uncertain or sections are missing. Use with score_coverage_ratio to QA extraction before trusting scores.