6.85 Review Topic Extraction, Sentiment Taxonomy & AI Summary Readiness
120 checks in Schema, Local, Trust, Deep Audit. AI-assisted
Scored in the Deep Audit's AI analysis pass.
Topic Extraction — Themes, Service Clusters & Intent Mapping
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.01 | Validate topics extracted from review corpus | Extracted | Heavy | Deterministic + AI |
| 6.85.02 | Validate primary topics align with core services | Aligned | Heavy | Deterministic + AI |
| 6.85.03 | Validate secondary topics captured (price, speed, communication) | Captured | Heavy | Deterministic + AI |
| 6.85.04 | Detect missing expected topics for the industry | None missing | Medium | Deterministic + AI |
| 6.85.05 | Validate topic extraction stable across time windows | Stable | Medium | Deterministic + AI |
| 6.85.06 | Validate topics grouped into coherent clusters | Coherent | Medium | Deterministic + AI |
| 6.85.07 | Validate topics map cleanly to user intents (quality, responsiveness, value, etc.) | Mapped | Medium | Deterministic + AI |
| 6.85.08 | Detect noisy or irrelevant topics (e.g., unrelated places/brands) | Minimal | Medium | Deterministic + AI |
| 6.85.09 | Validate topic labels human-interpretable | Interpretable | Medium | Deterministic + AI |
| 6.85.10 | Validate topic coverage representative of full review set | Representative | Medium | Deterministic + AI |
Sentiment Taxonomy — Per-Topic Positivity, Risk Clusters & Issue Themes
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.11 | Validate sentiment scored per topic cluster | Scored | Heavy | Deterministic + AI |
| 6.85.12 | Validate positive/neutral/negative split per topic computed | Computed | Heavy | Deterministic + AI |
| 6.85.13 | Validate high-risk negative topics identified (e.g., safety, fraud, damage) | Identified | Heavy | Deterministic + AI |
| 6.85.14 | Validate service-quality topics trend positive overall | Positive | Medium | Deterministic + AI |
| 6.85.15 | Detect topics with deteriorating sentiment over last 90 days | Minimal | Medium | Deterministic + AI |
| 6.85.16 | Validate conflict-resolution topics (support, refunds, callbacks) not skewed negative | Balanced | Medium | Deterministic + AI |
| 6.85.17 | Validate staff/technician mention topics largely positive | Positive | Medium | Deterministic + AI |
| 6.85.18 | Detect recurring negative subthemes (lateness, rudeness, overcharging) | Minimal | Medium | Deterministic + AI |
| 6.85.19 | Validate sentiment taxonomy stable across platforms (GBP, Yelp, FB, etc.) | Stable | Medium | Deterministic + AI |
| 6.85.20 | Validate per-topic sentiment scores stored for trend analysis | Stored | Medium | Deterministic + AI |
AI Summary Readiness — Topic Labeling, Summary Coherence & Risk Explanations
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.21 | Validate AI-generated topic labels match human-readable themes | Matching | Heavy | Deterministic + AI |
| 6.85.22 | Validate AI summaries reflect true sentiment distribution | Accurate | Heavy | Deterministic + AI |
| 6.85.23 | Validate AI can generate per-topic pros/cons lists from reviews | Generatable | Heavy | Deterministic + AI |
| 6.85.24 | Detect hallucinated topics not present in review corpus | None | Medium | Deterministic + AI |
| 6.85.25 | Validate summaries clearly differentiate topics (no blending of unrelated issues) | Clear | Medium | Deterministic + AI |
| 6.85.26 | Validate AI can surface top 3 positive themes per topic | Surfaceable | Medium | Deterministic + AI |
| 6.85.27 | Validate AI can surface top 3 recurring complaints per topic | Surfaceable | Medium | Deterministic + AI |
| 6.85.28 | Validate AI-generated narratives match real review language tone | Aligned | Medium | Deterministic + AI |
| 6.85.29 | Validate risk summaries highlight critical issues without exaggeration | Balanced | Medium | Deterministic + AI |
| 6.85.30 | Compute Review Topic & Sentiment Taxonomy Quality Score | Finalized | Critical | Deterministic + AI |
Language Patterns — Structure, Clarity & Naturalness
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.31 | Validate review language appears natural (non-AI, non-spam) | Natural | Heavy | Deterministic + AI |
| 6.85.32 | Detect repeated boilerplate language across multiple reviewers | None | Heavy | Deterministic + AI |
| 6.85.33 | Validate clarity of reviewer communication | Clear | Medium | Deterministic + AI |
| 6.85.34 | Detect overly generic phrasing (“great service,” “nice people”) | Minimal | Medium | Deterministic + AI |
| 6.85.35 | Validate grammar/spelling within normal variance | Normal | Medium | Deterministic + AI |
| 6.85.36 | Detect stylometric anomalies suggesting fake reviews | None | Medium | Deterministic + AI |
| 6.85.37 | Validate sentiment intensity matches wording (no mismatches) | Matching | Medium | Deterministic + AI |
| 6.85.38 | Validate language patterns consistent across demographic groups | Consistent | Medium | Deterministic + AI |
| 6.85.39 | Detect unusual linguistic clusters that may indicate coordinated review activity | None | Medium | Deterministic + AI |
| 6.85.40 | Validate the presence of specific, detail-rich narratives | Present | Medium | Deterministic + AI |
Recurring Themes — Positive & Negative Pattern Identification
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.41 | Validate recurring positive themes (quality, speed, value, professionalism) | Present | Heavy | Deterministic + AI |
| 6.85.42 | Detect recurring negative themes (delays, miscommunication, pricing issues) | Minimal | Heavy | Deterministic + AI |
| 6.85.43 | Validate emerging-new themes identified (e.g., new service lines, staff praise) | Identified | Medium | Deterministic + AI |
| 6.85.44 | Detect contradictory review themes across platforms | None | Medium | Deterministic + AI |
| 6.85.45 | Validate sentiment changes aligned with operational events (team change, policy change) | Aligned | Medium | Deterministic + AI |
| 6.85.46 | Detect “topic drift” — reviewers discussing unrelated subjects | None | Medium | Deterministic + AI |
| 6.85.47 | Validate themes properly categorized into service-relevant buckets | Categorized | Medium | Deterministic + AI |
| 6.85.48 | Detect highly polarized themes (extreme love/hate swings) | Low | Medium | Deterministic + AI |
| 6.85.49 | Validate consistency of experience themes (repeat mentions of same strengths/weaknesses) | Consistent | Medium | Deterministic + AI |
| 6.85.50 | Detect temporal clustering of negative themes (e.g., last 30 days anomaly) | Minimal | Medium | Deterministic + AI |
Linguistic Quality Signals — Trust, Authenticity & AI Interpretability
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.51 | Validate reviews include human-level detail & nuance | Detailed | Heavy | Deterministic + AI |
| 6.85.52 | Detect AI-generated review patterns (syntax repetition, unnatural flow) | None | Heavy | Deterministic + AI |
| 6.85.53 | Validate trust-enhancing linguistic markers (specific events, exact timing, names) | Present | Medium | Deterministic + AI |
| 6.85.54 | Detect emotionally manipulative patterns | None | Medium | Deterministic + AI |
| 6.85.55 | Validate readability of reviews (grade-level coherence) | Coherent | Medium | Deterministic + AI |
| 6.85.56 | Detect spam indicators (URLs, promos, competitor mentions) | None | Medium | Deterministic + AI |
| 6.85.57 | Validate linguistic patterns match typical customer demographics | Matching | Medium | Deterministic + AI |
| 6.85.58 | Validate AI can cleanly tokenize and classify linguistic patterns | Classifiable | Medium | Deterministic + AI |
| 6.85.59 | Validate no conflicting emotional cues within same review | None | Medium | Deterministic + AI |
| 6.85.60 | Compute Linguistic Quality & Recurring Theme Score | Finalized | Critical | Deterministic + AI |
Service-Line Mapping — Topics → Services → Business Units
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.61 | Validate review topics mapped to specific services (e.g., install, repair, consult) | Mapped | Heavy | Deterministic + AI |
| 6.85.62 | Validate each core service has sufficient review topic coverage | Sufficient | Heavy | Deterministic + AI |
| 6.85.63 | Validate positive themes distributed across main services (not just one) | Distributed | Heavy | Deterministic + AI |
| 6.85.64 | Detect services with disproportionately negative topic coverage | None | Medium | Deterministic + AI |
| 6.85.65 | Validate reviews reference correct service labels (no misclassification) | Correct | Medium | Deterministic + AI |
| 6.85.66 | Validate cross-sell/upsell topics (bundled services) appear in taxonomy | Present | Medium | Deterministic + AI |
| 6.85.67 | Detect missing topic coverage for newly launched services | None missing | Medium | Deterministic + AI |
| 6.85.68 | Validate branch/location-specific services reflected in topics | Reflected | Medium | Deterministic + AI |
| 6.85.69 | Validate topics distinguish between product issues vs. service issues | Distinct | Medium | Deterministic + AI |
| 6.85.70 | Validate service-line topic mapping stable over time | Stable | Medium | Deterministic + AI |
Operational Insights — Staff, Process, Pricing & Experience Themes
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.71 | Validate topics exist for staff behavior (friendliness, expertise, communication) | Present | Heavy | Deterministic + AI |
| 6.85.72 | Validate topics exist for operational reliability (on-time, scheduling, follow-through) | Present | Heavy | Deterministic + AI |
| 6.85.73 | Validate topics exist for pricing fairness/value | Present | Medium | Deterministic + AI |
| 6.85.74 | Detect recurring complaints about specific operational steps (booking, billing, wait time) | Minimal | Medium | Deterministic + AI |
| 6.85.75 | Validate topics highlight differentiators (speed, expertise, guarantees) | Highlighted | Medium | Deterministic + AI |
| 6.85.76 | Detect unresolved operational pain-point themes persisting over time | None | Medium | Deterministic + AI |
| 6.85.77 | Validate topics capture experience across full journey (pre-sale, delivery, post-sale) | Complete | Medium | Deterministic + AI |
| 6.85.78 | Detect themes indicating misalignment between marketing promises and actual delivery | None | Medium | Deterministic + AI |
| 6.85.79 | Validate operational improvements (if any) reflected in newer review topics | Reflected | Medium | Deterministic + AI |
| 6.85.80 | Validate topic taxonomy usable for internal CX/ops reporting | Usable | Medium | Deterministic + AI |
Risk Themes, Opportunity Clusters & AI-Ready Summarization
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.81 | Detect high-risk themes (safety, legal, fraud, severe negligence) | None | Heavy | Deterministic + AI |
| 6.85.82 | Detect reputation-risk clusters (discrimination, harassment, unethical conduct) | None | Heavy | Deterministic + AI |
| 6.85.83 | Detect compliance-related topics (contracts, disclosures, warranties) | Clean | Medium | Deterministic + AI |
| 6.85.84 | Validate positive “moat” themes (why customers choose this brand vs. others) | Present | Medium | Deterministic + AI |
| 6.85.85 | Validate opportunity clusters (features/services users wish existed) extracted | Extracted | Medium | Deterministic + AI |
| 6.85.86 | Validate AI-generated risk summary matches raw topic distribution | Matching | Medium | Deterministic + AI |
| 6.85.87 | Validate AI-generated opportunity summary matches topic signals | Matching | Medium | Deterministic + AI |
| 6.85.88 | Detect contradictions between AI summaries and underlying topics | None | Medium | Deterministic + AI |
| 6.85.89 | Validate topic & sentiment taxonomy exportable as structured data (JSON/CSV) | Exportable | Medium | Deterministic + AI |
| 6.85.90 | Compute Topic–Sentiment Operational Insight Master Score | Finalized | Critical | Deterministic + AI |
Cross-Platform Topic Alignment — Google, Yelp, Fb, Industry Sites
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.91 | Validate topic clusters consistent across all major review platforms | Consistent | Heavy | Deterministic + AI |
| 6.85.92 | Detect platform-specific discrepancies in sentiment or themes | Minimal | Heavy | Deterministic + AI |
| 6.85.93 | Validate each platform contains similar service-line coverage | Similar | Medium | Deterministic + AI |
| 6.85.94 | Detect mismatched narratives (positive on Google, negative on Yelp, etc.) | None | Medium | Deterministic + AI |
| 6.85.95 | Validate cross-platform entity naming consistent | Consistent | Medium | Deterministic + AI |
| 6.85.96 | Validate cross-platform metadata (categories/tags) align with topic distribution | Aligned | Medium | Deterministic + AI |
| 6.85.97 | Detect review sites with abnormal sentiment bias | None | Medium | Deterministic + AI |
| 6.85.98 | Validate platform-specific reviews still support global theme mapping | Supportive | Medium | Deterministic + AI |
| 6.85.99 | Validate multi-platform summaries align with unified taxonomy | Aligned | Medium | Deterministic + AI |
| 6.85.100 | Compute Cross-Platform Topic Alignment Score | Finalized | Critical | Deterministic + AI |
Multi-Language Review Analysis — Linguistic Consistency & NLP Equivalence
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.101 | Validate language detection correct for each review | Correct | Heavy | Deterministic + AI |
| 6.85.102 | Validate sentiment classification accurate across languages | Accurate | Heavy | Deterministic + AI |
| 6.85.103 | Validate multilingual terms map to correct topic clusters | Mapped | Medium | Deterministic + AI |
| 6.85.104 | Detect mistranslation risk or ambiguity in NLP parsing | None | Medium | Deterministic + AI |
| 6.85.105 | Validate code-switching (mixed languages in same review) handled correctly | Handled | Medium | Deterministic + AI |
| 6.85.106 | Validate non-English reviews show consistent themes with English ones | Consistent | Medium | Deterministic + AI |
| 6.85.107 | Detect sentiment drift caused by translation artifacts | None | Medium | Deterministic + AI |
| 6.85.108 | Validate language metadata usable for AI summarization | Usable | Medium | Deterministic + AI |
| 6.85.109 | Validate multi-language review sets well-represented in taxonomy | Represented | Medium | Deterministic + AI |
| 6.85.110 | Compute Multi-Language NLP Consistency Score | Finalized | Critical | Deterministic + AI |
Temporal Pattern Analysis — Time-Series Topic, Sentiment & Event Impact
| ID | Check | Passes when | Weight | Runs |
|---|
| 6.85.111 | Validate time-series topic stability | Stable | Heavy | Deterministic + AI |
| 6.85.112 | Detect sharp inflection points in topics (new issues emerging suddenly) | None | Heavy | Deterministic + AI |
| 6.85.113 | Validate sentiment evolution matches service improvements or declines | Matching | Medium | Deterministic + AI |
| 6.85.114 | Detect seasonal topic patterns (holidays, peak seasons, weather impact) | None or Expected | Medium | Deterministic + AI |
| 6.85.115 | Validate post-event sentiment recovery (after a negative incident) | Recovered | Medium | Deterministic + AI |
| 6.85.116 | Validate long-term sentiment trend upward or stable | Upward/Stable | Medium | Deterministic + AI |
| 6.85.117 | Detect short-lived anomalies that do not fit broader trends | None | Medium | Deterministic + AI |
| 6.85.118 | Validate topic recurrence intervals normal | Normal | Medium | Deterministic + AI |
| 6.85.119 | Validate time-series models (e.g., rolling averages) align with raw data | Aligned | Medium | Deterministic + AI |
| 6.85.120 | Compute Temporal Topic–Sentiment Stability Score | Finalized | Critical | Deterministic + AI |
← 6.84 External Link Graph, Referring Domains & Entity Voting Strength6.86 Business Hours, Coverage Reliability & Operational Signals →
Run these checks on your site
A Deep Audit scores every check here that applies to your page, then an AI pass of up to 150 checks. Included on Pro and Ultra, or $9 for one audit.
See plans
Machine-readable: catalog totals and the
full catalog as JSON (IDs, section, weight, status).