Personalized Search Ranking with Query and User Account Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engines face challenges in providing relevant search results due to ambiguity in search queries, vastness of information, language and cultural nuances, and complexities of personalization while respecting privacy, making it difficult to accurately interpret user intent and provide contextually relevant results.
Innovation Solution
A method involving a server computing device that generates query and user account vectors using transformer-based large language models, combines them to create an output vector, compares with item vectors to generate similarity scores, and orders results for personalized display, leveraging song metadata and audio features to enhance relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If search engines use sophisticated algorithms to interpret user intent, then search result relevance is improved, but system complexity increases
Solution Approach 1:
The patent introduces an intermediary vector representation layer between the search query and the search results. Query vectors and user account vectors serve as mediators that translate ambiguous natural language queries into mathematical representations that can be systematically compared with item vectors, thereby reducing the complexity of direct interpretation while improving relevance
Solution Approach 2:
The patent transforms the search problem from textual pattern matching to vector space manipulation. By converting queries and documents into vectors with specific dimensional parameters, the system can use mathematical operations (cosine similarity, dot products) to determine relevance, simplifying the algorithmic complexity while improving interpretability
2Quantity of substance
If search engines crawl and index vast amounts of data, then information coverage is improved, but resource consumption increases
Solution Approach 1:
The patent extracts only the essential features from the vast amount of indexed data and represents them as compact vector embeddings. Instead of processing entire documents, the system uses condensed vector representations that capture the most important semantic information, significantly reducing the computational resources needed while maintaining information coverage
Solution Approach 2:
The system performs preliminary vectorization of documents during the indexing phase, preparing compressed representations in advance. This preliminary action allows the search phase to operate much faster and with fewer resources, as it only needs to compare pre-computed vectors rather than process raw text data
3Reliability
If search engines personalize results based on user history, then user relevance is improved, but privacy concerns increase
Solution Approach 1:
The patent creates a simplified vector representation (user account vector) that captures user preferences without storing or exposing detailed personal history data. This copying approach allows the system to use user information for personalization while maintaining privacy, as the detailed historical data remains hidden and is only transformed into abstract vector representations
4Measurement precision
If search engines consider language and cultural nuances, then contextual accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent replaces complex linguistic processing mechanisms with vector space geometry. Instead of using complex NLP parsers and cultural knowledge bases, the system uses mathematical vector operations to capture language and cultural nuances, simplifying the processing while maintaining or improving contextual accuracy through the power of embedding representations
Data Source
AI summary
A technique that includes the steps of (1) generating a query vector based at least in part on a query received from a user account that is associated with a user account vector, (2) combining the query vector and the user account vector to establish a combined vector, (3) generating an output vector based at least in part on the combined vector, (4) obtaining, based on the query, a plurality of item vectors, where each item vector of the plurality of item vectors corresponds to a respective item, (5) comparing the output vector to the plurality of item vectors to generate respective similarity scores, (6) ordering the plurality of item vectors in accordance with the similarity scores to establish an ordered plurality of item vectors, and (7) causing the client computing device to display respective affordances for the items that correspond to the ordered plurality of item vectors.