LLM-Based Item Attribute Extraction for Catalog Search

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Unstructured item descriptions in online systems make it difficult to integrate data into other features, as the lack of structure prevents information from being usable for filtering or other system functionalities, limiting the utility of item attributes in search and recommendation systems.

Innovation Solution

An online system uses a machine-learned language model to extract item attributes from item descriptions by generating prompts, processing the output, standardizing the attributes, and storing them for use in item catalogs, enabling advanced search and recommendation systems.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If unstructured item descriptions are used in the item catalog, then the system can maintain simple data storage and processing, but the information cannot be used for filtering or other system features

Engineering Contradiction:

Improvedata storage simplicityVSAvoidsearch functionality

Core Design Contradiction:

Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The patent segments unstructured item descriptions into structured attributes by using a machine-learned language model to identify and extract specific product characteristics (e.g., color, size, material) from the text. This segmentation transforms the undifferentiated text into discrete, filterable attributes that can be independently used for search and filtering operations while maintaining the original description.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a machine-learned language model as an intermediary between the unstructured item descriptions and the structured data requirements of the online system. This intermediary automatically processes the text to extract meaningful attributes, bridging the gap between simple text storage and complex search functionality without requiring manual data entry or complex parsing systems.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If manual structuring of item descriptions is implemented, then search functionality is improved, but processing time and complexity increase

Engineering Contradiction:

Improvesearch functionalityVSAvoidprocessing time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent implements self-service by enabling the machine-learned language model to automatically extract and structure attributes from item descriptions without human intervention. The system autonomously processes incoming product data, identifies relevant attributes, and structures them for search functionality, eliminating the need for manual data entry or curation while maintaining high processing efficiency.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces manual mechanical processes (human analysts manually structuring data) with an automated machine-learned language model. This substitution uses artificial intelligence to perform the attribute extraction task that would otherwise require human time and effort, significantly reducing processing time while maintaining or improving the quality of structured data output.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Loss of information

If comprehensive attribute extraction is performed on all item descriptions, then data analytics capability is enhanced, but computational resources are consumed

Engineering Contradiction:

Improvedata analytics capabilityVSAvoidcomputational resource consumption

Core Design Contradiction:

Loss of informationVSUse of energy by moving object

Solution Approach 1:

The patent applies partial action by selectively extracting only the most relevant attributes from item descriptions based on the specific needs of different system features. Rather than processing every possible attribute uniformly, the system can adjust the depth and breadth of extraction depending on whether the goal is basic search functionality or advanced analytics, optimizing computational resource usage while maintaining data utility.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20240403947A1Extracting item attributes from item descriptions using large language machine-learned models

Publication Date: 2024.12.05 MAPLEBEAR INC

AI summary

A system may obtain an item description associated with an item in an item catalog. The system may generate a prompt for input to a machine-learned language model, the prompt specifying at least the item description and a request to identify one or more attributes of the item. The system may provide the prompt to a model serving system for execution by the machine-learned language model. The system may receive from the machine-learned language model, an output including a list of attributes and respective values associated with the item based on the item description. The system may standardize the formatting of the list of attributes and may store the list of attributes and the respective values for the list of attributes in association with the item in the item catalog.