Generates personalized web summaries and digest alerts by running custom search engines on a schedule and surfacing newly-matching content to the user, supporting daily-digest patterns over user-defined topical scopes.
Patent Overview
- Inventor
- Ramanathan V. Guha
- Assignee
- Google LLC
- Filed
- 2008-12-31
- Granted
- 2014-05-13
- Application Number
- US 12/347,801
The Challenge
The Challenge
Users with persistent topical interests benefit from regular updates rather than re-querying the same topic each day. The system needed a mechanism to run user-defined scoped queries on a schedule, summarize new matches, and deliver digests as alerts or summary streams.
- Re-Querying Is Repetitive Friction — Users interested in a topic must re-issue the same query daily to catch new content. The friction discourages sustained engagement with topics that matter.
- Topical Interests Are Persistent — Many user interests are stable: industries followed, brands tracked, topics studied. These persistent interests deserve standing-query treatment.
- Custom Search Engines Scope Queries — CSEs already define topical scope through specification. Scheduling them as standing queries produces topical-digest streams without new infrastructure.
- Summaries Must Be Concise — Digest content must summarize, not just list. Users opening an alert want the gist, not all results. Summarization compresses while preserving signal.
- Delivery Cadence Must Match User Preference — Some users want hourly alerts; some want daily digests; some want weekly summaries. The system must support configurable cadence.
Innovation
How The System Works
The patent schedules custom search engines as standing queries, detects new content matching the CSE specification since the last run, summarizes the new matches, and delivers the summary to the user as an alert or digest on the user's configured cadence.
- User Defines CSE And Cadence — The user creates a CSE (or subscribes to an existing one) and specifies digest cadence: real-time, hourly, daily, weekly. Subscription preferences save per user.
- Schedule CSE As Standing Query — The standing-query scheduler runs the CSE on its configured cadence. Each run scopes to content not yet seen by this user.
- Detect New Matches — Compared to the last run's match set, identify new content matching the CSE specification. New matches are the digest candidates.
- Score And Filter For Quality — Per match, quality and relevance scoring applies. Low-scoring matches are dropped; top matches enter the digest.
- Summarize Content — Per match, generate a concise summary. Summaries combine title, key sentences, and entity highlights into digest-ready form.
- Compose Digest — The digest combines summarized matches into a coherent format: top items first, grouped by sub-topic if applicable. Layout adapts to delivery surface (email, push notification, in-app feed).
- Deliver To User — Digest delivers via the user's chosen channel and cadence. Engagement feedback (open rate, click-through) refines future digest composition.
Custom Search As Standing Query
The patent's load-bearing idea is to turn user-defined custom search engines into standing queries that produce digest streams. The CSE specification becomes the topical scope; the scheduler runs it continuously; the summarizer produces user-ready output.
Persistent Interests Become Continuous Streams
Topical interests that persist deserve continuous treatment, not repeated one-shot queries. Standing queries plus summarization plus delivery cadence produce a topical-digest experience users return to over time.
- Scheduled Standing Queries — CSEs run on a schedule. Each run detects new content since the last run; the user gets only fresh material.
- Summarization Layer — Per match, summarization compresses content into digest-ready form. Users scan summaries; click for full content.
- Configurable Cadence — Per user, delivery cadence configures: real-time, hourly, daily, weekly. The system supports the rhythm each user prefers.
Technical Foundation
Technical Foundation
The patent specifies the user subscription store, the standing-query scheduler, the new-match detector, the quality filter, the summarizer, the digest composer, and the multi-channel delivery layer.
- User Subscription Store — Per user, stores CSE subscriptions, configured cadence, and last-run timestamps. Subscriptions can be added, paused, or removed by the user.
- Standing Query Scheduler — Runs CSEs on configured cadence. Scheduler distributes load to avoid spike at common cadence boundaries (daily-9am).
- New Match Detector — Per run, detects content matching the CSE specification that did not match in prior runs. Tracks per-user-per-CSE seen-set.
- Quality Filter — Per new match, quality and relevance scoring filters. Below-threshold matches are excluded from the digest.
- Summarization Layer — Per match, summarization extracts key sentences and entity highlights. Output is digest-ready content of bounded length.
- Multi-Channel Delivery — Digests deliver via email, push notification, in-app feed, or other channels per user preference. Format adapts to channel.
The Process
The Process
The pipeline runs as a continuous stream of scheduled runs producing user-tailored digests. Latency from new content publication to digest delivery depends on cadence.
- User Creates Subscription — User selects CSE and configures cadence. Subscription saves to the store.
- Scheduler Triggers Run — Per user-CSE-cadence, scheduler triggers a run. Load distributes across the day.
- Detect New Matches — Match detector identifies content matching CSE specification since last run for this user.
- Filter For Quality — Quality filter excludes low-scoring matches. Top matches proceed.
- Summarize Matches — Per match, summarizer produces digest-ready content.
- Compose And Deliver — Digest composes; delivery layer pushes via the user's chosen channel.
- Capture Engagement — Open, click, and dismiss events log. Engagement refines future digest quality scoring and composition.
Quality Control
Quality Control
Bad digests cause unsubscribes. The patent specifies safeguards.
- Quality Floor — Below-threshold matches are excluded. Better to send a smaller digest than to dilute with weak content.
- Frequency Tuning — If a CSE generates excessive volume, the system can suggest broader cadence (hourly to daily). Avoids overwhelming users.
- Disengagement Backoff — Users who consistently ignore digests get reduced delivery. Per-user backoff respects expressed disinterest.
- Summary Quality Audit — Summaries are audited for accuracy and faithfulness. Bad summaries break trust.
- Easy Unsubscribe — One-click unsubscribe in every digest. User control is first-class.
Real-World Application
Customized digest primitives underpin Google Alerts, the topic-following surfaces in Discover, and various email-digest features across Google products. The pattern of standing-query plus summarization plus delivery is foundational.
- Standing-query Execution Model — CSEs run on a schedule as standing queries. Standing-query treatment is the digest pattern.
- Summarization Compression Layer — Per match, summarization compresses to digest-ready form. Users scan rather than read full content in the digest.
- Configurable cadence User Control — Per user, cadence configures. The system supports many rhythms.
Why Email Digests Drive Repeat Engagement
Sustained engagement on topics depends on automatic delivery rather than user-initiated re-querying. Digests produce repeat exposure that compounds engagement over time.
Why Topic Following Is The Engagement Substrate
Surfaces like Discover and topic-following features build on the same primitive: persistent user interest becomes continuous content delivery. Sites covering topics consistently win sustained visibility through these channels.
<\/section>What This Means for SEO
What This Means for SEO
The patent turns user-defined custom search engines into standing queries that run on a schedule, detect newly-matching content, summarize it, and deliver digests. SEO implication: persistent topical interest becomes a continuous delivery channel, so consistent publishing on a topic earns repeat exposure without the user re-querying.
- New Content Is The Trigger — The system detects content matching the standing query since the last run. Publishing fresh, clearly-scoped content on a topic is what surfaces you into digests and alerts. A static page that never updates rarely re-enters the new-match stream.
- Tight Topical Scope Matches Standing Queries — Custom search engines define a topical scope, and content that fits that scope cleanly gets surfaced. Pages with unambiguous topical focus match standing queries more reliably than broad-spectrum content that fits any one scope only weakly.
- Summarizable Structure Helps Delivery — The pipeline summarizes new matches for the digest. Content with clear headlines, lead sentences, and extractable structure summarizes cleanly and presents better in the delivered digest than dense undifferentiated prose.
- Cadence Builds Repeat Exposure — Digests deliver on a recurring cadence, producing repeated touchpoints the user did not initiate. Consistent publishing rhythm aligns you with that cadence, compounding exposure over time rather than spiking once and disappearing.
- Topic Following Is The Same Substrate — Surfaces like Discover and topic-following build on this standing-query primitive. Sites that cover topics consistently win sustained visibility through these follow channels, which bypass per-query ranking entirely.
- Earn The Standing-Query Match Once, Benefit Repeatedly — Once your content reliably matches a topical standing query, you surface on every refresh that finds new qualifying material from you. Establishing topical authority on a scope is high-leverage because it pays out across every scheduled run.
- Engagement Reinforces Future Delivery — Digest interaction feeds back into what gets delivered. Content that earns opens and clicks in a digest strengthens future selection. Optimizing the digest-facing snippet and headline matters as much as the page body for sustained delivery.