Query Suggestion System Using Local Index and Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
E-commerce platforms face inefficiencies in generating query suggestions, leading to high latency and reduced user engagement due to reliance on multiple third-party microservices, resulting in suboptimal performance and throughput.
Innovation Solution
A system that generates query suggestions by utilizing a database with a query index and control circuit to process query prefixes, retrieving and scoring candidates on a single computing device, incorporating decay functions for temporal scoring and channel-specific scores, and executing processes locally to reduce latency and enhance performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple third-party microservices are used to generate query suggestions, then query suggestion functionality is provided, but latency increases and performance decreases
Solution Approach 1:
The patent consolidates multiple third-party microservices into a single integrated system on one computing device. The query suggestion system combines query processing, scoring, and suggestion generation functions that were previously distributed across multiple external services, eliminating inter-service communication overhead and reducing latency from 225ms to 25ms.
Solution Approach 2:
The patent introduces an intermediary query index structure that caches and pre-processes query data locally. This intermediary layer between the user interface and the suggestion generation logic enables fast retrieval of query suggestions without requiring real-time communication with multiple external microservices, significantly reducing response time.
2Reliability
If multiple third-party microservices are used to generate query suggestions, then query suggestion functionality is provided, but throughput decreases
Solution Approach 1:
The patent merges multiple microservice functions into a single high-performance system that can process queries locally without the overhead of inter-service communication. This consolidation enables the system to handle 89% more query suggestions per second compared to the distributed microservice architecture.
3Adaptability or versatility
If query suggestions are generated using external microservices, then comprehensive query processing is achieved, but system complexity increases
Solution Approach 1:
The patent combines multiple independent microservices into a single integrated system, reducing architectural complexity from multiple distributed components to one unified service. This merger maintains comprehensive query processing capabilities while eliminating the complexity of service orchestration, authentication, and inter-service communication protocols.
4Measurement precision
If decay functions and channel-specific scores are incorporated, then query suggestion accuracy is improved, but calculation complexity increases
Solution Approach 1:
The patent pre-calculates and stores decay function results and channel-specific scores in the query index during system initialization or periodic updates. This preliminary action allows the system to retrieve pre-computed scores during query processing without performing complex real-time calculations, maintaining high accuracy while reducing runtime computational complexity.
Data Source
AI summary
In some embodiments, apparatuses and methods are provided herein useful to generating query suggestions. In some embodiments, a system for generating query suggestions for retail comprises a database comprising a query index including queries and query scores, and a control circuit configured to receive a query prefix from a user interface of a retail channel, retrieve, from the query index and based on the query prefix, query correction candidates, calculate edit distances for each of the query correction candidates, select, based on the edit distances, one or more query suggestions from the query correction candidates, calculate, based on the edit distances, a prefix completion cost for each of the query suggestions, calculate, based on the prefix completion costs and the query scores, the query correction costs for each of the query suggestions, and sort, based on the query correction costs, the query suggestions.