An approach to search that enables users to find results based on relationships between multiple entity types, transforming how we discover information through spatial and temporal connections.
Patent Overview
- Filed
- November 27, 2020
- Granted
- September 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 with Traditional Search — Traditional search engines excel at simple queries with single criteria, such as "Starbucks near San Francisco Airport" or "Films shot during World War II." However, they struggle with more complex, compositional queries that involve relationships between multiple entity...
- Computational Challenge — The time complexity of the basic algorithm is O(N×M), where N represents the number of entity references of the first type and M represents the number of entity references of the second type. For large datasets with 10,000 entities of each type, this requires 100,000,000...
- Rigid Structure — Traditional approaches require users to specify exact entities or locations, rather than allowing for relative relationships between multiple entity types.
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.
- Key Innovation — This patent introduces a technique for providing search results based on compositional queries - searches that involve relationships between multiple entity types rather than single, fixed criteria. Unlike traditional search engines that handle simple...
- Understanding Compositional Queries — A compositional query is a search that includes at least two types of entity references related by a relative relationship. This represents a fundamental shift from traditional search paradigms.
- Performance Benefits — This approach dramatically reduces computation time by: Instead of performing 100 million calculations for 10,000×10,000 entity pairs, the system can reference pre-computed nearest neighbors.
- Limited Scope — Existing systems typically solve queries based on a single, fixed location criterion or a single, fixed temporal criterion, limiting the complexity of searches users can perform.
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.
- Data Structure Foundation — The system leverages a sophisticated knowledge graph - a data structure organized as nodes and edges representing entities and their relationships. Key Components: Each node in the knowledge graph contains unique identification references, allowing the...
- Network Infrastructure — Internet, local networks, and communication protocols (TCP/IP, Wi-Fi, cellular) enable data transmission between user devices and server systems.
- Search Engine Servers — Process compositional queries, access the knowledge graph, perform entity identification and comparison, and generate result sets. Multiple servers may work in parallel for scalability.
- Database Servers — Store the knowledge graph data structure, pre-generated tables, and indexed content. May be distributed across multiple servers and geographic locations for redundancy and performance.
- Infrastructure Planning — Analyzing spatial relationships between critical facilities, such as nuclear reactors near geological hazards or hospitals near population centers.
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.
- Result Generation — One or more resultant entity references are determined based on the comparison, representing the best matches for the compositional query.
- Result Presentation — Qualifying nuclear reactors and volcanoes are displayed on an interactive map with visual indicators, allowing users to explore the spatial relationships.
- Computational Challenges — Processing relationships between multiple entity types requires comparing potentially thousands or millions of entity pairs, creating significant computational overhead.
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.
- Complex Criteria — The system identifies entity types, determines relationships, and compares attribute values to find the best matching pairs or groups of entities.
- Entity Type Identification — The system determines the first entity type, second entity type, and relationship type from the compositional query through natural language processing and semantic analysis.
Real-World Application
The patent shapes how the search engine behaves in production. These are the visible outcomes for users and content publishers.
- 100 Practical Example: Nuclear Reactors and Volcanoes — Consider the compositional query: "Nuclear reactors within 100 miles of volcanoes." This demonstrates the system's capability to process complex spatial relationships.
- Spatial Relationships — "American Banks close to Japanese restaurants" - involves two place types with a relative spatial relationship, without specifying which specific restaurant or bank.
- Temporal Relationships — "Companies that went bankrupt during an economic crisis" - involves two event types with a relative time relationship.
What This Means for SEO
What This Means for SEO
Compositional queries combine entities and constraints, and pages that expose those combinations explicitly capture them.
- Pages That Combine Constraints Win Long-Tail — A query like "vegan pizza near me open now" combines multiple constraints. Pages that explicitly cover each combination (vegan, location, hours) capture the long tail.
- Filterable Content Captures More Variations — A listing page with explicit filters becomes the canonical answer for many composed queries. Each filter is an indexable entry surface if the URL changes.
- Schema For Each Composition Layer — Mark up the entities, attributes, and constraints separately. The system composes the answer from the structured pieces, not from the prose blob.