Machine Learning UoM Normalization for Ecommerce Search Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Ecommerce systems face challenges due to variations in units of measure (UoM) representations, leading to ambiguities and errors in procurement, higher return rates, and lower conversions.
Innovation Solution
The use of machine learning to optimize UoM representations through a UoM representation recommender model, which normalizes UoM representations in search queries and product catalogs, and generates new attribute names and taxonomies based on contextual data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple measurement systems (SI, US, imperial) are supported to increase adaptability, then system versatility improves, but representation ambiguity increases
Solution Approach 1:
The patent introduces a machine learning model as an intermediary that automatically selects and standardizes unit representations based on product category and context. This mediator resolves the conflict by translating between different measurement systems into a standardized internal representation, maintaining versatility while eliminating ambiguity.
Solution Approach 2:
The system dynamically changes the representation parameters of units based on the product category and contextual information. By adjusting which unit system is used (SI, US, or imperial) according to the specific product type, the system maintains adaptability across different markets while ensuring consistent, unambiguous representations within each category.
2Ease of operation
If non-standard UoM representations are allowed in product catalogs to maintain ease of operation, then data entry flexibility improves, but search accuracy deteriorates
Solution Approach 1:
The system performs preliminary standardization of unit representations during product catalog ingestion and indexing, before search operations occur. The machine learning model pre-processes and normalizes unit data based on product categories, so that when searches are performed, the data is already in a standardized, searchable format without requiring manual intervention.
Solution Approach 2:
The patent replaces manual standardization processes with an automated machine learning-based normalization system. Instead of relying on operators to consistently apply standard unit representations, the system automatically detects and standardizes units using learned patterns from product data, maintaining ease of data entry while ensuring search accuracy.
3Reliability
If manual standardization of UoM representations is implemented to improve reliability, then representation consistency improves, but processing time increases
Solution Approach 1:
The system implements self-service standardization where the machine learning model automatically normalizes unit representations without requiring manual intervention. The model learns from existing product catalog data and autonomously applies appropriate unit standardizations based on product categories, achieving representation consistency while eliminating the time cost of manual processing.
Solution Approach 2:
The patent substitutes manual standardization operations with an automated machine learning-based normalization process. The system uses trained models to automatically detect, interpret, and standardize unit representations, replacing time-consuming manual efforts with rapid automated processing that maintains high representation consistency.
4Measurement precision
If UoM normalization is applied to all product attributes to improve search precision, then search accuracy improves, but computational complexity increases
Solution Approach 1:
The system applies unit normalization selectively based on product category and attribute type rather than uniformly to all attributes. The machine learning model identifies which product attributes require normalization and applies appropriate standards only to those cases, reducing unnecessary computational overhead while maintaining search accuracy where it matters most.
Solution Approach 2:
The system dynamically adjusts the normalization parameters and intensity based on product category characteristics. By changing which normalization rules are applied and with what strictness based on the specific product type and attribute, the system optimizes the balance between search precision and computational complexity for different domains.
Data Source
AI summary
Methods and systems are provided for using machine learning to optimize UoM representations. In embodiments described herein, units of measure (UoMs) and relationships of each of UoMs to textual representations of each of the UoMs are stored in a knowledge graph. Text corresponding to a measurement of a product is extracted by an inference model. A recommended textual representation of the measurement of the product by is determined by an autoencoder model including a corresponding textual representation of one of the UoMs from the textual representations of the one of the UoMs stored in the knowledge graph. The recommended textual representation of the measurement of the product is then displayed.