Product Search System Using Image and Text Vector Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing product search technologies using image-based queries often fail to accurately match user intentions due to reliance on image similarity alone, leading to inappropriate product suggestions, especially when text information is incomplete or irrelevant.
Innovation Solution
A method and electronic device system that processes both image and text information by generating vectors from recognized text in the query image and product information, prioritizing search results based on similarities in both image and text features to provide a more accurate product search result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If only image similarity is used for product search, then the search process is simple and fast, but the search accuracy deteriorates and completely different products may be presented
Solution Approach 1:
The patent combines image similarity matching and text information matching into a unified product search system. The server performs both image-based similarity comparison and text extraction/comparison, then integrates these two dimensions to determine final product priorities, thereby improving search accuracy while maintaining operational efficiency
Solution Approach 2:
The patent introduces text information as an intermediary element that mediates between the query image and product database. The text extracted from images serves as a bridge to enhance matching accuracy, allowing the system to filter and prioritize products more effectively beyond simple image similarity
2Measurement precision
If text information is extracted from images to improve search accuracy, then product matching improves, but incomplete or irrelevant text extraction may lead to inappropriate results
Solution Approach 1:
The system implements feedback mechanisms where the server evaluates both image similarity and text matching results, then uses this combined feedback to adjust and determine final product priorities. This feedback loop ensures that incomplete or irrelevant text extraction does not compromise overall result reliability, as the system can compensate through image similarity analysis
Solution Approach 2:
The patent performs both image similarity matching and text information matching, even though text extraction alone may be insufficient. By implementing both actions (image matching and text matching), the system ensures that even if text extraction is incomplete or irrelevant, the overall search reliability is maintained through the complementary image-based approach
3Measurement precision
If both image and text information are processed, then search accuracy improves, but the system complexity increases
Solution Approach 1:
The server is designed with multi-functionality, performing both image processing and text extraction/processing within a single unified system. This universal approach allows the system to handle multiple types of data (image and text) through one platform, improving search accuracy while avoiding the need for separate complex systems for each function
Data Source
AI summary
The disclosure is for displaying a search result for products. A server may include a communication module, a memory including a database which stores product information in association with one or more product images corresponding to a product and text information related to the product, and a processor. The processor may receive a query image for an external object from an electronic device by using the communication module, select, from the database, a plurality of pieces of product information corresponding to the external object, based on similarities between the one or more product images and the query image, extract the text information for the external object included in the query image, determine a priority among the plurality of pieces of product information, based on a comparison between the extracted text information and the text information included in the selected plurality of pieces of product information, and transmit, to the electronic device, at least some of the plurality of pieces of product information and information on the priority, in response to the reception, by using the communication module.