LLM-Based Item Combination Ranking for Online Concierge Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Customers find it difficult and time-consuming to compare items and retailers in online concierge systems, especially when looking for the best combination of items within a specific timeframe and budget.
Innovation Solution
An online concierge system that determines and ranks combinations of items for retailers based on a free-text query, using a large language model to extract item categories and constraints, and then computes scores for each combination based on user and item data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If customers manually compare items and retailers across multiple options, then they can find the best combination of items within their constraints, but the process becomes time-consuming and difficult
Solution Approach 1:
The system performs preliminary actions by automatically generating item combinations and computing scores for all retailers before the customer needs to compare. The backend service pre-processes item data, generates combinations based on constraints, and computes scores in advance, so when the customer views results, the comparative analysis is already complete and ready for immediate selection.
Solution Approach 2:
The system introduces an intermediary component (the combination generation and scoring service) between the customer's comparison needs and the raw item data. This intermediary automatically generates item combinations, computes scores based on multiple criteria, and presents processed comparison results, eliminating the need for customers to manually compare items while maintaining high precision.
2Adaptability or versatility
If the system provides detailed item combinations from multiple retailers, then customers get comprehensive options, but the system complexity increases
Solution Approach 1:
The system segments the complex task of item comparison into distinct modular components: constraint extraction from user queries, item combination generation, score computation for each combination, and result presentation. Each module handles a specific aspect independently, making the overall system manageable despite its versatility in handling multiple retailers and constraints.
Solution Approach 2:
The combination generation service is designed as a universal system that can handle multiple retailers, various item types, and different constraint conditions simultaneously. Rather than creating separate systems for each retailer or constraint type, a single multi-functional service processes all scenarios using the same core logic, reducing overall system complexity while maintaining high adaptability.
3Measurement precision
If the system computes scores for item combinations based on multiple constraints, then the results are more accurate and useful, but the computational effort increases
Solution Approach 1:
The system performs preliminary computation by pre-generating all possible item combinations and pre-computing their scores based on constraints before the customer needs to make a decision. This upfront computational effort, though energy-intensive, is done once in the background, allowing fast retrieval and comparison when the customer interacts with the system.
Solution Approach 2:
The system computes scores for all possible item combinations across all retailers, even though the customer will ultimately select only one combination. This excessive computation ensures that no potentially optimal combination is missed, achieving maximum scoring accuracy at the cost of additional computational energy that is justified by the quality of results.
Data Source
AI summary
An online concierge system receives a free-text query describing items and constraints from a client device associated with a user. The system generates a prompt including the query and a request to identify the items and constraints. The system provides the prompt to a large language model, extracts, from an output of the model, the constraints and one or more categories associated with the items, and identifies retailers based on user data associated with the user. For each retailer, the system identifies a set of items associated with each category, determines, based on the constraints, a combination of a subset of items associated with each category, and computes a score for the combination based on the user data and item data associated with items in the combination. The system ranks the combinations based on the scores and sends information describing a ranked set of the combinations to the client device.