Patent US10303684, filed by Google and granted on May 28, 2019, discusses methods and systems for enhancing search engine rankings by modifying resource scores based on “entity selection values” associated with search terms.
In the context of information retrieval, this patent outlines a method to leverage user interactions, such as clicks and selections, to refine search rankings, particularly for resources that have limited historical interaction data.
This adjustment methodology enhances relevance and accuracy in search results, especially for new or lesser-known resources.
Key Components of the Patent
Resource Scoring Adjustment System
The system comprises several interconnected components: a network that includes publisher sites, a search engine, user devices, and two specialized sub-systems—a Search Term-Entity Selection Evaluator and a Search Score Adjuster.
These components collaborate to collect, analyze, and adjust search scores based on interactions (clicks or selections) of resources associated with specific entities (such as a person, place, or thing) mentioned in search terms.
Search Term-Entity Selection Evaluator
This component calculates “selection values” for entities within resources. It does this by evaluating historical data from users’ search selections. Essentially, it assesses how often users select a resource in response to specific search terms.
The selection evaluator uses search term data and resource data to create a metric known as the Search Term-Entity Selection Value, representing user interest in certain entities.
Search Score Adjuster
Using the selection values calculated by the evaluator, the search score adjuster modifies the ranking of resources. It mainly adjusts rankings for resources that reference entities with high selection values, promoting resources that align better with user interests.
The score adjuster is especially beneficial for new resources with minimal interaction data, allowing them to be promoted based on associated entities’ popularity.
Detailed Mechanism of Resource Scoring Adjustment
- Data Collection and Analysis
- The system accesses resource data from various resources on the internet, including details about the entities each resource references.
- It then collects search term data, which includes a selection value for each resource based on user interactions in response to search terms.
- Calculation of Search Term-Entity Selection Values
- For each search term and each referenced entity, a selection value is calculated. This value is based on how frequently users select resources that mention the entity when presented in search results.
- For example, if a search term like “Mount Fuji” leads users to click on resources related to entities like “Fuji apples” and “Mount Fuji” itself, these clicks are quantified as selection values.
- Weighting and Aggregation
- Resources with multiple entities are assigned weight scores for each entity to reflect the significance of each within the content.
- The evaluator uses these weights and selection values to compute a composite score for each entity based on how often it is associated with user-selected resources.
- Adjusting Rankings Based on Entity Selection Values
- When a user submits a search query, the search engine uses historical data and the selection values to prioritize resources that reference popular entities.
- By factoring in selection values related to entities, the score adjuster can boost new resources associated with high-demand entities, even when they lack a substantial click history.
Illustrative Example: Entity-Based Scoring in Action
Consider three resources that reference entities “Fuji apples,” “Mount Fuji,” and “Keiko Fuji.”
- The system calculates how often these entities are selected based on user interactions with each resource in response to search terms like “Fuji,” “climb Fuji,” or “Fuji singer.”
- If “Mount Fuji” is frequently selected in response to “climb Fuji,” resources referencing “Mount Fuji” receive a scoring boost.
- This adjustment allows the search engine to prioritize resources based on entity selection relevance, even if those resources have limited historical data.
Applications and Benefits
- Enhanced Search Accuracy
- The patent’s methodology improves search engine accuracy by aligning search results more closely with user intent as indicated by past selection behavior.
- Support for New Resources
- By leveraging entity selection values, the system can elevate new resources based on their association with popular entities, helping them appear more frequently in relevant search results.
- Broad Application Across Various Entity Types
- The system can adapt to a wide array of entities, such as locations, products, or individuals, making it versatile across multiple domains of search.
Potential Challenges and Considerations
- Sparse Data for Rare Entities: For entities with limited interaction data, selection values might be lower, potentially affecting ranking adjustments.
- Privacy Concerns: Collecting user selection data raises privacy issues, but anonymization and data aggregation can mitigate these risks.
- Relevance Over Time: Entity selection trends might change, necessitating regular updates to the entity selection values to maintain ranking accuracy.
SEO Takeaways
- Entity-Centric Content Optimization
- Emphasize and clearly define entities in your content that are likely to be linked to popular search terms. For example, if your content references “Mount Fuji,” explicitly mention related terms like “climb,” “tourism,” or “Fuji apples” to increase relevance.
- New Content Strategies
- For new pages or sites with little historical traffic, focus on associating content with popular entities. This can help search engines recognize your content’s relevance even if user engagement data is limited.
- Optimize for Long-Tail Keywords with Strong Entity Associations
- Using long-tail keywords (like “climb Mount Fuji in autumn”) that closely match user intent can help boost the relevance of your page in search results due to higher selection values tied to specific entities.
- Leverage Multi-Entity Content to Cover Broader Search Interests
- Content that references multiple, relevant entities (e.g., combining “Mount Fuji” and “Keiko Fuji” in a cultural guide) can potentially benefit from cross-entity selection values, improving the chances of appearing for various related queries.
- Regularly Update Content to Maintain High Entity Selection Values
- As user interests shift, refreshing content to stay aligned with trending entities can sustain or enhance search rankings based on entity relevance over time.
- Keyword Research Focused on Entity Significance
- When conducting keyword research, consider terms associated with high-interest entities. This approach aligns content with potential entity selection values, which may result in better visibility for new or updated pages.
By integrating entity-based optimization strategies informed by the mechanisms in this patent, SEO efforts can better align with Google’s sophisticated ranking adjustments, ultimately enhancing visibility and search performance.
System and Method for Providing Search Results
Inventors: David C. Orr, Daniel E. Loreto, and Michael J. P. Collins
Assignee: Google LLC
US Patent: 8,346,757
Granted: January 1, 2013
Filed: August 27, 2013
