Classifies landing pages using trained classifiers based on search queries. Determines appropriate page types for web searchers — the search-to-landing-page intent bridge in advertising contexts.
Patent Overview
- Inventor
- Andrei Broder
- Assignee
- Yahoo! Inc.
- Filed
- 2008-12-23
- Granted
- 2014-11-11
The Challenge
The Challenge
Ad landing pages must match user intent. A query implies a likely landing-page type (product page, comparison, review, definition). Classifying landing pages and matching to query-intent type produces better ad experience.
- Wrong Landing Page Type Wastes Click — Click lands on wrong page type — user leaves immediately. Wasted impression, wasted click.
- Queries Imply Landing-Page Type — Different queries imply different appropriate landing-page types.
- Landing-Page Classification Must Generalize — Per page, classification must work across many domains and page styles.
- Classification Plus Query-Intent Match — Both sides must be classified for matching to work.
- Match Quality Drives Ad Performance — Better matching produces better ad-quality scores and better user experience.
Innovation
How The System Works
The system trains landing-page classifiers, classifies each candidate landing page, classifies queries by sought page-type, matches landing-page classification to query-intent type, and uses match quality in ad selection and ranking.
- Train Landing-Page Classifiers — Labeled examples train per-type classifiers (product, comparison, review, definition, etc.).
- Classify Landing Pages — Per ad landing page, assigned one or more type labels.
- Classify Queries By Sought Type — Per query, infer sought landing-page type.
- Match Types — Per (query, landing-page) pair, type alignment scored.
- Combine With Other Ad Signals — Type match combines with bid, quality, relevance signals.
- Select And Rank Ads — Per query, top-matched ads selected and ranked.
- Learn From Engagement — Click and post-click engagement feeds back into matching.
Landing-Page Type Is The Match Dimension
The patent's load-bearing idea is that landing-page type is a first-class ad-matching dimension. Beyond keyword match, type-match between query intent and landing-page format determines ad-experience quality.
Type Alignment Drives Quality
Per (query, landing-page) pair, type alignment is the quality signal. Misaligned types waste clicks; aligned types deliver value to user and advertiser.
- Landing-Page Classifiers — Per type, learned classifiers identify landing-page formats.
- Query-Type Inference — Per query, sought landing-page type inferred.
- Type-Match Scoring — Per pair, type alignment scored.
Technical Foundation
Technical Foundation
The patent specifies the classifier trainer, page classifier, query-type inferrer, matcher, ad-selection integrator, and learning loop.
- Classifier Trainer — Labeled examples train per-type classifiers.
- Page Classifier — Per landing page, assigns type labels.
- Query-Type Inferrer — Per query, infers sought page type.
- Matcher — Per pair, scores type alignment.
- Ad-Selection Integrator — Type match combines with other ad signals.
- Learning Loop — Engagement feeds back into matching.
The Process
The Process
Classification runs offline and per page; matching runs per query.
- Train Classifiers — Offline, classifiers trained.
- Classify Pages — Per page, type labels assigned.
- Receive Query — Query arrives.
- Infer Query Type — Per query, sought type inferred.
- Match Types — Per pair, alignment scored.
- Select And Rank Ads — Top matches selected.
- Learn From Engagement — Feedback loops.
Quality Control
Quality Control
Wrong classification corrupts matching. The patent specifies safeguards.
- Classifier Validation — Per type, classifier validated against labeled data.
- Query-Type Calibration — Per query type, inference calibrated.
- Engagement Feedback — Mis-matched pairs detected via post-click engagement.
- Adversarial Defense — Pages misrepresenting their type flagged.
- Continuous Retraining — Classifiers retrain against fresh data.
Real-World Application
Context-transfer in search advertising is foundational for landing-page-quality scoring across ad platforms. The pattern of landing-page-type classification plus query-type matching underpins modern ad-quality scores.
- Per-page Classification Granularity — Each landing page receives type labels.
- Per-query Intent Inference — Each query carries sought-type inference.
- Type-aligned Match Criterion — Type alignment drives match quality.
Why Clear Page-Type Signals Compound Ad Quality
Per landing page, clear type signals (clean product page, clear comparison structure, evident definition format) earn correct classification and better type-match scoring with intent-aligned queries.
Why Mismatched Bidding Wastes Budget
Bidding on queries whose intended type doesn't match your landing-page type wastes budget. Matching landing-page type to query-intent type is the structural quality optimization.
<\/section>What This Means for SEO
What This Means for SEO
This patent classifies landing pages by type, infers the page type a query is seeking, and scores ad quality on the alignment between them. SEO implication: matching page format to query intent is a first-class quality dimension, not just keyword relevance.
- Page Type Is A Ranking Dimension — The system classifies whether a page is a product page, comparison, review, or definition, then matches it to the type a query wants. Format alignment with intent is scored directly, beyond keyword match.
- Match Landing Format To Query Intent — A transactional query seeking a product page that lands on a definition page is a type mismatch that wastes the click. Building the page format the query actually wants is the structural optimization.
- Clear Format Signals Earn Correct Classification — A clean product page, an evident comparison table, or a clearly structured definition gets classified correctly. Ambiguous, do-everything pages classify poorly and match weakly.
- Engagement Feedback Catches Mismatches — Post-click engagement feeds back into matching, so pages that pull clicks but cause immediate exits are detected as mismatched. Genuine intent satisfaction, not just attracting the click, is what holds up.
- Type Misrepresentation Is Flagged — Pages that misrepresent their type are flagged by the system. Dressing a thin page up as a comprehensive comparison to win a match invites adversarial detection.
- The Principle Generalizes To Organic — Landing-page-type quality scoring became the backbone of modern quality scores. The same logic, serving the format the query expects, transfers to building organic pages that satisfy intent type.
- One Page, One Intent Type — Because classification assigns a type, pages that try to serve several intent types at once dilute their type signal. Focused pages match their intended query type more cleanly.