Search Engine Item Option Preselection via Deep Learning Clustering

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional search engines face challenges in providing accurate and efficient search results due to the vast number of potential matches for a search query, making it difficult for users to find contextually relevant webpages or items, especially when there is limited data available about the user's intent.

Innovation Solution

Employing deep learning methods to identify item options by training a machine learned model using user features and purchase history, which predicts item options based on the common user history of similar users, and presenting these options as preselections on the search results page, thereby condensing search results and reducing the need for users to sift through numerous options.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If conventional search engines return all potential search results, then the completeness of search results is improved, but the complexity of the search results page and user difficulty in finding relevant information increases

Engineering Contradiction:
Improvenumber of search resultsVSAvoidsearch results page complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent extracts and presents only the most relevant item options (such as color, storage, model) from the complete set of search results. The machine learned model identifies and extracts key item options that users are most likely to be interested in, displaying them prominently while excluding less relevant options, thus reducing page complexity while maintaining result completeness.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the search results by categorizing items into different item option categories (e.g., color, storage, model). Instead of presenting all options in a single undifferentiated list, the results are segmented into organized groups with preselected options based on user preferences and purchase history, making the information more manageable and easier to navigate.

Inventive Principle:
Principle #1Segmentation

2Loss of information

If conventional search engines present all item options, then the completeness of information is improved, but the time required for users to find desired items increases

Engineering Contradiction:
Improveinformation completenessVSAvoiduser search time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The patent performs preliminary action by using the machine learned model to preselect the most likely item options before the user even views the search results. Based on user features and purchase history, the system预先 identifies and highlights the options the user is most likely to choose, allowing users to quickly find desired items without manually filtering through all options.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent replaces the mechanical system of manual user filtering and selection with an automated intelligent system. The machine learned model automatically analyzes user data, predicts preferences, and presents optimized search results, substituting the manual exploration process with an automated recommendation system that reduces user search time while maintaining information completeness.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If search engines process more search requests with the same resources, then the productivity is improved, but the accuracy of predicting user intent decreases due to limited data

Engineering Contradiction:
Improvesearch processing capacityVSAvoiduser intent prediction accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary mechanism - the machine learned model trained on user features and purchase history - that bridges the gap between limited user data and accurate intent prediction. This intermediary system processes and analyzes user data more efficiently, enabling accurate predictions even with limited information, thus allowing the system to handle more search requests without sacrificing prediction accuracy.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4113330A1Item option identification and search result presentation at a search engine
Publication Date: 2023.01.04 EBAY INC
  • EP4113330A1 patent drawingFigure 1
  • EP4113330A1 patent drawingFigure 2
  • EP4113330A1 patent drawingFigure 3

AI summary

A recommendation engine utilizes deep learning methods, including machine learned neural network models, to identify user group clusters of users having a common purchase history when determining item options, such as an item feature, for a user associated with a search query at a search engine. The determined item options are presented to the user at a search results page or as an item listing as a preselection of selectable options for item options of an item option category, thereby identifying and providing a specific item variation. The search results page can be condensed by excluding items having a same set of item option categories or a same identified item option, thereby providing a search results page or item listing that allows other contextually relevant items to be provided and identified by the user.