17579337. EYES-ON ANALYSIS RESULTS FOR IMPROVING SEARCH QUALITY simplified abstract (MICROSOFT TECHNOLOGY LICENSING, LLC)

EYES-ON ANALYSIS RESULTS FOR IMPROVING SEARCH QUALITY

Organization Name

MICROSOFT TECHNOLOGY LICENSING, LLC

Inventor(s)

Xi Yun of Redmond WA (US)

Jiantao Sun of Bellevue WA (US)

Zheng Chen of Bellevue WA (US)

Kaushik Chakrabarti of Bellevue WA (US)

Leon Melvin Romaniuk of Snohomish WA (US)

Pingjun Hu of Sammamish WA (US)

Mingyu Wang of Issaquah WA (US)

Wei Li of Bellevue WA (US)

Yaxi Li of Bellevue WA (US)

Abhilash Srivastava of Bellevue WA (US)

EYES-ON ANALYSIS RESULTS FOR IMPROVING SEARCH QUALITY - A simplified explanation of the abstract

This abstract first appeared for US patent application 17579337 titled 'EYES-ON ANALYSIS RESULTS FOR IMPROVING SEARCH QUALITY

Simplified Explanation

The patent application describes systems that generate and use training data to train learn-to-rank models while preserving the privacy of client data. Here are the key points:

• The systems extract features and patterns from user queries, search results, and user interactions without tracking, storing, or transmitting the actual user data.
• The systems infer the quality of search results based on user behavior data and optionally query intentions.
• The systems generate and label training data based on the inferred search result quality.
• This training data is used to train learn-to-rank models, which improve the search quality of new user queries that have similar features and patterns as the filtered and labeled training data.

Potential applications of this technology:

• Improving search engines: The learn-to-rank models trained using this approach can enhance the search quality of search engines by considering user behavior and query intentions.
• Personalized recommendations: By analyzing user interactions and search patterns, the systems can generate personalized recommendations for users without compromising their privacy.
• Ad targeting: The inferred search result quality can be used to optimize ad targeting, ensuring that relevant ads are shown to users.

Problems solved by this technology:

• Privacy preservation: By not tracking, storing, or transmitting user data, the systems address privacy concerns associated with using personal information for training models.
• Enhanced search quality: The learn-to-rank models trained using this approach can improve the relevance and accuracy of search results, leading to a better user experience.
• Efficient training: By generating and labeling training data based on inferred search result quality, the systems can train the models effectively without relying on actual user data.

Benefits of this technology:

• Privacy protection: The systems ensure that user data is not compromised or exposed during the training process.
• Improved search relevance: By considering user behavior and query intentions, the learn-to-rank models can deliver more relevant search results.
• Efficient and scalable training: The approach of generating and labeling training data based on inferred search result quality allows for efficient and scalable training of the models.

Original Abstract Submitted

Systems are configured for generating and utilizing training data to train learn-to-rank type models in a manner that preserves privacy of client data used for generating the training data. The systems extract features and patterns of the user queries, search results and user interactions with the search results without tracking, storing or transmitting underlying values of the user data to preserve privacy of the user data. Systems are also configured to infer search result quality based on at least the user behavior data, and optionally query intentions, and to generate and label corresponding training data accordingly. This training data is applied to learn-to-rank type models to train the learn-to-rank type model to improve search quality of search results provided by the learn-to-rank type models when new user queries are processed that having features and patterns corresponding to the filtered and labelled training data.