Dynamic Question Recommendation for Retail Procurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In retail procurement, buyers face challenges in associating relevant questions with categories and objects, leading to missed questions and slowed data creation processes, especially when new objects are added to the master data.
Innovation Solution
A machine learned model is used to identify relevant questions based on a category hierarchy and object features, automatically generating questions for new objects by filtering out irrelevant ones and determining applicability based on attributes and prerequisites.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If buyers manually associate questions with categories and objects, then question relevance can be ensured, but the process becomes time-consuming and questions may be missed
Solution Approach 1:
The system enables automatic question generation where the software autonomously identifies and associates relevant questions with categories and objects using machine learning models, eliminating the need for manual buyer intervention while maintaining high relevance through automated feature analysis and probability scoring
Solution Approach 2:
The system pre-generates questions and their associated features in advance, storing them in a database ready for rapid retrieval and assignment when new objects are added, thereby preparing question associations before they are actually needed rather than creating them on-demand
2Reliability
If all basic questions are added manually for each new object, then completeness can be achieved, but the data creation process slows down significantly
Solution Approach 1:
The system dynamically adjusts question selection based on object parameters and features by feeding object attributes into machine learning models that output probability scores, automatically filtering to retain only the most relevant questions rather than applying a fixed set of all possible questions
Solution Approach 2:
The system retrieves and adapts pre-existing question templates and their associated features from the database based on category hierarchies and object similarities, copying proven question structures rather than creating questions from scratch for each new object
3Adaptability or versatility
If a large category and object base is maintained, then system versatility improves, but associating appropriate questions becomes more difficult
Solution Approach 1:
The system segments the large category and object base into hierarchical groups and processes questions at different levels (category-level inherited questions and object-specific questions), breaking down the complex association task into manageable segments that can be handled systematically through machine learning feature matching
Solution Approach 2:
The system creates a universal question template structure with reusable features that can apply across multiple categories and objects, where a single question template can serve multiple purposes through parameter substitution rather than requiring unique questions for each object
Data Source
AI summary
In an example embodiment, a request is received, via a graphical user interface, to add a new object to a directory of objects, the new object having a first category in a hierarchical taxonomy of categories and objects. Then one or more questions previously assigned to the first category and/or one or more existing objects within the first category are retrieved. Each of the retrieved one or more questions and information about the new object are then fed into a first machine learned model trained to output a probability that a question is applicable to an object. One or more questions are generated for the new object based on the probability for each of the retrieved one or more questions. At least one of the one or more generated questions is then assigned to the new object.