Product Identification via ML Attribute Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current e-commerce systems lack robustness in identifying specific products based on product model numbers, leading to frustration for users who cannot find exact matches, as these systems rely on simple string-matching and fail to recognize proprietary or exclusive model numbers, resulting in inadequate search results and alternative product recommendations.
Innovation Solution
A computer-implemented system using machine-learning algorithms to analyze user queries, extract attributes, and determine product categories, enabling the identification of target products and recommending alternative products based on pattern recognition and experimental data, even if exact matches are not found in the database.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If simple string-matching is used to identify products, then the system is easy to implement and fast, but it fails to recognize proprietary or exclusive model numbers not in the database
Solution Approach 1:
The patent segments the product model number into multiple attributes (e.g., brand, series, specifications) and uses machine learning to independently evaluate each segment. This allows the system to identify products even when the complete model number is not in the database, as long as sufficient attribute segments match known patterns.
Solution Approach 2:
The patent introduces an intermediary machine learning model that acts as a bridge between the user's query and the product database. Instead of direct string matching, the ML model translates the query into attribute-based representations, enabling recognition of proprietary model numbers through inferred attributes rather than exact database matches.
2Measurement precision
If exact model number matching is required, then search precision is high, but user experience deteriorates when exact matches are not found
Solution Approach 1:
The patent applies partial matching by evaluating individual attributes of the model number separately rather than requiring complete exact matches. The machine learning model can identify products based on partial attribute matches (e.g., brand and series without exact specifications), providing sufficient precision while improving user experience through flexible matching.
Solution Approach 2:
The patent changes the matching parameters from exact string equality to probabilistic attribute-based scoring. The machine learning model evaluates multiple attributes with different weights and generates a confidence score, allowing the system to maintain high precision through multi-attribute verification while improving ease of operation through flexible, non-exact matching.
3Reliability
If machine learning algorithms are used to extract attributes and identify products, then product identification robustness improves, but computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary action by pre-training the machine learning model with extensive product data and attribute patterns before deployment. During actual product identification, the pre-trained model quickly processes queries using learned patterns, reducing real-time computational complexity while maintaining high robustness in product identification.
Solution Approach 2:
The patent substitutes traditional mechanical string-matching algorithms with machine learning-based attribute extraction and evaluation. The ML approach replaces rigid character-by-character comparison with intelligent pattern recognition, achieving superior robustness in handling proprietary model numbers while the pre-trained model keeps computational overhead manageable.
4Ease of operation
If alternative product recommendations are generated, then user experience improves when exact matches are not found, but system complexity increases
Solution Approach 1:
The patent implements feedback by using the attribute extraction and evaluation framework for both exact product identification and alternative product recommendation. The same machine learning model that identifies the target product also ranks alternative products based on attribute similarity, providing a unified approach that improves user experience without requiring separate complex recommendation systems.
Data Source
AI summary
Disclosed embodiments provide systems and methods for identifying a target product and generating alternative product recommendations based on a user query. A computer-implemented system may be configured to perform operations comprising using machine learning to determine a plurality of attributes and at least one pattern associated with a user's product model number search query. The operations may further comprise determining at least one queried product of interest by the user and at least one product category based on an experimental data set. The operations may further comprise determining a target product based on the queried product of interest. The operations may further comprise determining a plurality of key features associated with the queried product based on experimental data, and determining at least one top alternative product. The operations may further comprise transmitting the target product and the top alternative product for display on an external device to the user.