An approach to search engine optimization that automatically generates intelligent query filters by analyzing resource content, user behavior, and semantic relationships, transforming how users discover and refine information.
Patent Overview
- Granted
- October 2023
The Challenge
The Challenge
The problem this patent addresses comes from limits in how earlier systems handled the underlying signal. Several specific gaps motivated the new approach.
- The Problem — Search queries are often incomplete expressions of user informational needs. Traditional search engines rely on hardcoded filters requiring expert knowledge and manual programming, which cannot adapt to the dynamic nature of web content. Users frequently refine searches after...
- Adaptive Precision — Filters tailored to specific user needs at the time of search and available results, enabling domain-specific refinement.
- The Operating Environment — The system operates within a comprehensive network ecosystem connecting publisher websites, user devices, and the search engine through computer networks including LANs, WANs, and the Internet.
Innovation
How The System Works
The patent introduces a multi-step mechanism that turns the input signal into a usable ranking output. Each step builds on the previous one.
- Language Model Integration — Language models facilitate sophisticated query-similarity findings and semantic understanding, enabling the system to identify relationships between terms that may not be immediately obvious.
- Key Innovation Pillars — Patent US11797626B2 - Methods, systems, and apparatus for providing filters from resource content. Inventors: Ian MacGillivray, Kaylin Spitz, Selena Sunling Yang, Varun Jasjit Singh, Emma S. Persky, Yonatan Erez. Assignee: Google LLC.
- Dynamic Learning — Filters automatically learned offline and generated at serving time, dramatically improving system performance and saving human effort.
- Content-Driven Intelligence — Learns from any relevant metadata or text, including reviews, descriptions, and user interactions to create contextually appropriate filters.
Technical Foundation
Technical Foundation
The implementation rests on a specific set of components and data structures. These are the parts the patent claims and the engineering that ties them together.
- Search Engine Core — Crawls and indexes resources, processes queries, and generates intelligent search results with dynamically created filters.
- Technical Implementation Architecture — The system can be implemented in digital electronic circuitry, computer software, firmware, or hardware, including various structural configurations and combinations.
- System Performance and Scalability — The system is designed to handle thousands of publisher websites and user devices simultaneously, processing queries and generating filters in real-time or near-real-time.
- Computing Infrastructure — System can be implemented across distributed computing infrastructures including web services, distributed computing, and grid computing models. Components interconnected through various communication networks including LANs, WANs, and the Internet...
The Process
The Process
In production, the system executes a sequence of stages from query reception to result delivery. Each stage applies one transformation to the data.
- Query Submission — Users submit search queries through web browsers or applications, including unique identifiers from cookies or user accounts.
- Filter Generation Pipeline — The filter generation process transforms raw resource content into intelligent, user-selectable query refinement options through a sophisticated multi-stage pipeline.
- Filter Selection Process — When users select one or more query filters, the system receives the selection and provides a filtered set of content results. The filtered set is a proper subset of the unfiltered results, showing only content matching both the original query and selected...
- Query Reception — System receives search query with one or more terms (words, numbers, symbols). Process invoked when query is categorical, highly indicative of particular category.
Quality Control
Quality Control
The system includes checks that defend against edge cases, manipulation, and degraded signal. Without these, the core mechanism would be exploitable.
- Filter Generation — Filter subsystem analyzes resource content to identify contextually relevant query filters for the identified resources.
- Candidate Filter Determination — The system determines candidate filters through sophisticated analysis of query relationships and term relevance, creating a proper subset of the initial keyword set.
- Quality Scoring and Term Prominence — Candidate filters are rated based on multiple criteria to ensure the most relevant and useful filters are presented to users.
- Location-Based Quality — Candidate filters appearing in prominent resource positions (titles, headings) receive higher quality scores than those in metadata or body text.
- User Interface and Filter Display — Query filters are displayed in user interfaces within web browsers or applications capable of providing query features, presented in search results pages.
Real-World Application
The patent shapes how the search engine behaves in production. These are the visible outcomes for users and content publishers.
- Universal Application — Can be learned from any relevant metadata or text, enabling application across diverse domains from restaurant reviews to product descriptions to academic papers.
- Broad Application Domains — While described in the context of general search engines, these features can be applied to any system or application that searches a data store, including specialized corpus-specific applications.
- Publisher Websites — Thousands of websites hosting resources including HTML pages, documents, images, videos, and audio files with embedded metadata and hyperlinks.
What This Means for SEO
What This Means for SEO
When result filters are auto-generated from page content, the attributes you expose become the filter slots you can occupy.
- Filter Slots Are Content-Derived — The filters in a SERP come from the structured attributes the system has found in indexed pages. Exposing more attributes opens more filter slots.
- Filtered Result Pages Are Indexable Surfaces — When a filter generates a stable URL, that URL is an indexable surface in its own right. Build crawlable filtered listings, not just JavaScript-driven ones.
- Attribute Naming Standardization Wins — When all pages in a category use the same attribute names, the filter is robust. Diverging attribute names fragment the filter and cost everyone visibility.