Crowd Assisted Query System for Visual Item Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The vast amount of electronically stored data in networked systems, often with inconsistent naming conventions, makes it difficult to find specific items using standard search methods.
Innovation Solution
A crowd assisted query system that uses a graphical user interface to receive queries with data objects, including media content and text, and allows users to provide description suggestions, with a reward system to incentivize accurate suggestions, which are then used to train an artificial intelligence module for improved item identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard search methods are used to find items in networked systems, then the search process is simple and direct, but the search accuracy deteriorates due to inconsistent naming conventions and vast amounts of electronic data
Solution Approach 1:
The patent introduces an intermediary image matching system between the user and the item database. Instead of directly searching text descriptions with inconsistent naming conventions, the system uses image-based visual search as a mediator to identify items, bypassing the text inconsistency problem while maintaining search simplicity for users.
Solution Approach 2:
The patent replaces the traditional text-based mechanical search system with an image-based visual search system. By substituting text processing with image recognition and matching, the system achieves higher search accuracy despite the complexity of implementing image processing infrastructure.
2Measurement precision
If crowd sourcing is implemented to collect description suggestions, then the accuracy of item identification improves, but the system complexity and operational overhead increase
Solution Approach 1:
The patent implements a self-service crowd sourcing mechanism where users voluntarily contribute item descriptions and images to the database in exchange for rewards or recognition. The system automatically processes and integrates these contributions without requiring manual curation, maintaining operational simplicity while improving item identification accuracy through diverse user inputs.
Solution Approach 2:
The patent incorporates feedback loops where the system evaluates crowd-sourced description suggestions and provides reinforcement to contributors based on the quality and usefulness of their inputs. This automated feedback mechanism maintains system simplicity while continuously improving item identification accuracy through learned patterns from crowd contributions.
3Productivity
If an AI module is trained using crowd-sourced data, then the productivity of item search improves, but the time and resources required for training increase
Solution Approach 1:
The patent performs preliminary actions by continuously collecting and pre-processing crowd-sourced images and descriptions in the background, preparing training data in advance. This allows the AI module to be trained more efficiently when deployment is needed, as the data preparation work has already been accomplished, reducing the perceived training time and improving search productivity.
Solution Approach 2:
The patent implements continuous data collection and model improvement, where the AI module is continuously trained on new crowd-sourced data rather than undergoing periodic lengthy retraining. This continuous learning approach maintains high search productivity while distributing the training time burden, making the system more responsive and efficient over time.
Data Source
AI summary
Aspects of the present disclosure relate to a network-based crowd assisted query system that includes a client device in communication with an application server executing the crowd assisted query system over a network. For example, the crowd assisted query system may be or include a group of one or more server machines. Users of the crowd assisted query system are presented with a graphical user interface (GUI) configured to receive queries that include data objects, wherein the data objects include representations of unidentified items of interest to the user. The data objects may include media content, such as graphical images as well as audio data, and in some example embodiments may further include text data describing the unidentified items.