Attribute Inconsistency Detection Using Structured and Unstructured Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems fail to reliably detect inconsistencies between unstructured and structured attributes in item listings, leading to customer confusion and increased costs due to incorrect item information display.
Innovation Solution
A machine learning model, such as a deep neural network, is used to analyze unstructured and structured attributes for item listings, detecting inconsistencies and generating alerts for corrections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional systems are used to display item listings, then the system structure remains simple, but inconsistencies between unstructured and structured attributes are not detected
Solution Approach 1:
A machine learning model is introduced as an intermediary component between the unstructured and structured attribute data. This model receives both types of attributes as input and outputs a consistency determination, acting as a mediator that detects inconsistencies without requiring complex integration of the attributes into a single structured format.
Solution Approach 2:
The patent replaces manual or rule-based consistency checking mechanisms with a machine learning-based detection system. The ML model learns patterns and relationships between unstructured and structured attributes, substituting complex manual verification processes with an automated intelligent system that can handle diverse attribute formats.
2Reliability
If manual inconsistency detection methods are used, then detection accuracy improves, but time and resources required increase significantly
Solution Approach 1:
The machine learning model performs self-service by automatically learning from training data the relationships and patterns between unstructured and structured attributes. Once trained, the model can independently detect inconsistencies in new data without requiring manual intervention or complex rule sets, enabling rapid and accurate detection.
Solution Approach 2:
The system changes the approach from manual rule-based parameter checking to data-driven parameter learning. The ML model learns optimal parameters and decision boundaries from training examples, allowing it to adapt to various attribute formats and detect inconsistencies efficiently without requiring explicit programming of all possible inconsistency scenarios.
Data Source
AI summary
Systems and methods are provided for detecting inconsistencies between product attributes, in which product attributes may be divided into structured and unstructured attributes. Structured and unstructured attributes are represented in text, and analyzed by a model architecture such that an alert may be generated when an inconsistency between a related element of the structured attributes and unstructured attributes is detected. In order to detect an attribute inconsistency, an input vector is formed using the structured attributes, unstructured attributes, and additional semantic information enabling the comparison (e.g., “Long” and “L” are equivalent).