An intelligent product configuration method and system based on multi-level semantic matching

By employing a multi-level semantic matching method, combined with vector retrieval, semantic reordering, and a large language model, the efficiency and accuracy issues of non-standard product configuration in the engineering industry are resolved. This achieves efficient and reliable intelligent product matching and decision-making, applicable to intelligent matching and selection scenarios for various products.

CN122286326APending Publication Date: 2026-06-26GUANGZHOU BAOLUN ELECTRONICS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGZHOU BAOLUN ELECTRONICS CO LTD
Filing Date
2026-03-19
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing technologies struggle to efficiently and accurately handle non-standard product configuration requirements in engineering industries. This is especially true when there are numerous product types and complex parameters, resulting in low configuration efficiency, poor consistency, and a lack of effective semantic normalization capabilities. Consequently, configuration results lack clear decision-making basis and fail to meet the needs of manual review and business auditing.

Method used

A multi-level semantic matching method is adopted, which achieves intelligent product matching through vector retrieval, semantic re-ranking, and joint decision-making by a large language model. First, vectorization encoding is performed through a pre-set text embedding model, and approximate nearest neighbor retrieval is performed using a product vector index library to form a candidate product set; then, fine-grained relevance modeling is performed through a semantic re-ranking model, and finally, a large language model performs joint evaluation of multiple candidates and outputs a structured result.

Benefits of technology

It enables efficient recall and accurate matching in a large-scale product database, reduces adaptation costs, improves the reliability, traceability and overall efficiency of configuration decisions, and outputs reliable and standardized configuration results that are easy for manual review and business auditing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122286326A_ABST
    Figure CN122286326A_ABST
Patent Text Reader

Abstract

This invention discloses an intelligent product configuration method and system based on multi-level semantic matching, comprising: acquiring a demand dataset, extracting feature information from demand items and concatenating them to form demand text; using a text embedding model for vectorized encoding, performing an approximate nearest neighbor search in a product vector index library to obtain a first candidate product set; pairing the demand text with the feature text of the first candidate products, inputting it into a semantic re-ranking model to complete relevance modeling and outputting a matching score, and re-ranking and filtering based on the score to obtain a second candidate product set; inputting the demand text and the product information of the second candidate product set into a large language model, which performs multi-candidate joint evaluation and outputs a structured evaluation result; after parsing and verifying the structured evaluation result, generating and outputting a configuration result containing the selected product identifier. This invention achieves multi-level semantic matching and intelligent product configuration, improving matching accuracy and decision reliability.
Need to check novelty before this filing date? Find Prior Art