Author: Olaf Kopp
Reading time: 6 Minutes

Recent interest based relevance scoring

Topics: , ,

Rate this post

The patent describes a computer-implemented method for processing query information that involves receiving previous and current queries within a certain period from a search requester. It includes receiving search results based on the current query, each identifying a document associated with a relevance score to the query. The method involves determining a category based on prior queries and identifying previous activity periods of other searchers whose queries match the current query and indicate the same category. It aims to adjust the relevance of search results by comparing the selection statistics of these matched queries from other searchers to general selection statistics, thus generating adjusted scores for search result documents based on their relevance to the current query and the identified category.

  • Patent Number: US 9390143 B2
  • Date of Patent: July 12, 2016
  • Inventors: Philip A. McDonnell (San Francisco, CA), Glen M. Jeh (San Francisco, CA), Taher H. Haveliwala (Fremont, CA), Yair Kurzion (Mountain View, CA)
  • Assignee: Google Inc., Mountain View, CA (US)
  • Filed: January 22, 2015
  • Countries: USA


The background of the patent delves into the challenges of enhancing the relevance of search results provided by search engines. It highlights the ongoing effort to better understand what users are seeking when they input queries and how to present the most pertinent documents or items in response. This involves making inferences based on various clues about the user’s needs, which can be specific to the individual or more general.

One approach to determining relevance involves analyzing the links between documents; for example, a document linked by many other relevant documents might be deemed particularly pertinent. This concept is based on the assumption that web authors link to pages they believe are useful for their audience, thus a page with many such “votes” is likely to be of high quality.

The background also discusses the use of user reactions to search results as an indicator of relevance. The idea is that if users frequently select a particular search result, it likely means that result is relevant to their query. Thus, monitoring and analyzing which search results users click on can provide valuable feedback for improving search algorithms.

Additionally, the background mentions the potential for inferring user needs from additional information, such as the user’s language or location. This information can complement direct feedback from user interactions with search results, enabling search engines to refine their relevance algorithms further.

Overall, the background of this patent outlines the complexity of accurately interpreting search queries and the various strategies employed to improve the relevance of search results. It sets the stage for introducing the patent’s specific contribution to this field, which focuses on leveraging recent interest and activity data to refine search relevance.


The patent document US 9390143 B2 encompasses claims focused on a method for enhancing the relevance of search results based on the recent interests and activities of users. Here’s a summary of the key claims outlined in the patent:

  • Computer-Implemented Method for Query Processing: This involves receiving a sequence of queries (prior and current) within a specific activity period from a search requester. The method includes analyzing these queries to determine relevance by examining a plurality of search results each associated with a document and a score indicating its relevance to the current query.
  • Determining Categories Based on Prior Queries: The method includes identifying categories or interests based on the analysis of prior queries submitted by the user.

1. Storing Recent Activity

      • The system captures and stores data related to the recent search queries and interactions of users. This includes the queries submitted, the search results clicked, the duration spent on the resultant pages, and any subsequent queries related to the same topic or interest area.

2. Analyzing Prior Queries

      • The stored data is analyzed to identify patterns that indicate the user’s areas of interest. This involves examining the content and context of the queries and the user’s interaction with the search results. Machine learning algorithms or heuristic methods could be used to classify these queries into specific interest categories.

3. Determining Categories Based on Analysis

      • Based on the analysis, the system categorizes the queries into predefined interest categories. These categories could range broadly from sports and entertainment to more niche topics like quantum computing. The categorization process takes into account the semantic meaning of the queries and the nature of the content that users engage with.

4. Identifying Similar User Activities

      • The system then identifies other users who have shown similar search behaviors or interests, based on their recent search activities and the categories of their interests. This step involves comparing the current user’s categorized interests with the stored profiles or recent activities of other users to find matches.

5. Adjusting Search Result Relevance

      • For future queries, the search engine uses the identified interest categories and the behaviors of similarly interested users to adjust the relevance of search results. This could mean promoting or demoting certain results in the rankings based on the presumed relevance to the user’s recent interests.

6. Dynamic Categorization and Relevance Adjustment

      • The categorization and relevance adjustment process is dynamic and continuously updated based on new user activities. As users’ interests evolve, the system recalibrates the categories they’re associated with and adjusts search results accordingly to ensure continued relevance.

  • Utilizing Prior Activity Periods: Identifying previous activity periods of other search requesters whose queries match the current query and share the same category of interest. This is used to adjust the relevance of search results for the current query.

  • Adjusting Search Results Based on Prior Activity: The method involves adjusting the relevance scores of search results based on the activities and selections made by users in the identified category during their prior activity periods. This includes analyzing both the selection statistics of these users and general selection statistics to generate adjusted scores for the search result documents.
  • Ranking Adjustments: Based on the adjusted scores, the method includes ranking the search result documents accordingly for the current query, taking into account the newly calculated relevance scores that reflect the user’s recent interests and activities.
  • Dynamic Adaptation: The claims further encompass dynamically adapting the ranking of search results in real-time or near-real-time based on the ongoing analysis of user activities and the detected categories of interest, ensuring that the search results remain relevant to the user’s current informational needs.

The essence of these claims lies in creating a more personalized and dynamic search experience by leveraging data on users’ recent search activities and interests. This approach aims to improve the accuracy and relevance of search results, making them more tailored to the specific needs and preferences of each user at the moment of their search.

The principles outlined in the patent are widely applicable in search engine technologies, where personalization and relevance are crucial for improving user experience. The methodology could potentially be applied in scenarios such as:

  • Personalized Search Engines: Tailoring search results for individual users based on their recent search history and clicked links to provide more relevant information.
  • E-commerce Platforms: Adjusting product recommendations and search results based on the browsing and purchasing history of shoppers to better match their interests.
  • Content Discovery Platforms: Curating news feeds or content suggestions in social media or news aggregation services based on users’ interaction with similar content in the past.

Implications for SEO

  • User Behavior Analysis: SEO strategies might need to evolve beyond traditional keyword optimization and metadata tweaks. Understanding and analyzing user behavior, including search history and interaction with search results, becomes crucial. This means SEO efforts could benefit from tools and techniques that track and analyze how users interact with content related to specific queries.
  • Personalization of Content: Given the patent’s focus on adjusting search results based on user interests, SEO strategies may need to consider how to make content more personalized or tailored to different user segments. This could involve creating diverse content that appeals to different interests or user behaviors identified through data analysis.
  • Diversification of SEO Tactics: Relying solely on traditional SEO techniques may not suffice. Incorporating user experience (UX) optimization, enhancing site navigation based on user interest patterns, and creating content that addresses the specific needs or questions of various user groups could become increasingly important.
  • Engagement Metrics: With search engines potentially using data on user interactions (e.g., click-through rates, time spent on pages) to adjust rankings, focusing on engagement metrics might become more critical. SEO strategies could include optimizing content to increase engagement and reduce bounce rates, thereby signaling relevance and value to search engines.
  • Semantic Search Optimization: As the patent involves analyzing user queries and interests to adjust search relevance, there’s an implication that semantic search optimization will play a bigger role. This means optimizing content for topic clusters, user intent, and contextual relevance, not just specific keywords.
  • Cross-Platform User Interest Data: Considering user interests might involve analyzing data across different platforms (e.g., social media, forums, other web properties). SEO strategies could look into how brand presence and content strategy on various platforms influence search behavior and, consequently, search rankings on major search engines.


Content from the blog

LLMO: How do you optimize for the answers of generative AI systems?

As more and more people prefer to ask ChatGPT rather than Google when searching for read more

What is the Google Knowledge Vault? How it works?

The Google Knowledge Vault was a project by Google that aimed to create an extensive read more

What is BM25?

BM25 is a popular ranking function used in information retrieval systems to estimate the relevance read more

The dimensions of the Google ranking

The ranking factors at Google have become more and more multidimensional and diverse over the read more

Interesting Google patents for search and SEO in 2024

In this article I would like to contribute to archiving well-founded knowledge from Google patents read more

What is the Google Shopping Graph and how does it work?

The Google Shopping Graph is an advanced, dynamic data structure developed by Google to enhance read more