Interactive Learning Classifiers With Human Feedback for Large Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing systems struggle to efficiently classify and understand large volumes of unstructured data due to the limitations of machine learning, requiring extensive labeled data and expert intervention, while humans can provide semantic understanding but are inefficient at scale.

Innovation Solution

An interactive system that integrates human feedback with machine learning, allowing users to select features and labels through an intuitive interface, leveraging both human semantic understanding and computational scalability to improve classification accuracy and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If machine learning techniques are used for data classification, then classification speed and scalability are improved, but classification accuracy and semantic understanding deteriorate due to lack of labeled data

Engineering Contradiction:
Improveclassification speedVSAvoidclassification accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system implements an interactive feedback loop where users provide feedback on classified items, and the machine learning model uses this feedback to iteratively improve its classification accuracy. The feedback mechanism allows the system to learn from user corrections and refine its understanding of data patterns over time.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system introduces an intermediary layer between automated classification and final data labeling. This intermediary presents classified data to users for review and correction, serving as a bridge that combines machine speed with human semantic understanding. The intermediary feedback mechanism enables continuous improvement of the classification model.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If extensive labeled data is collected for machine learning, then classification accuracy is improved, but time and resource consumption increase

Engineering Contradiction:
Improveclassification accuracyVSAvoiddata preparation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary classification of large datasets using automated machine learning techniques before human review is needed. This preliminary action filters and organizes data in advance, reducing the amount of data that requires manual processing and accelerating the overall data preparation timeline.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The feedback mechanism allows users to correct misclassifications efficiently, and the system learns from these corrections to improve future classifications. This feedback loop reduces the need for extensive manual labeling by automatically incorporating user insights into the model's learning process.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If human experts manually label and analyze data, then semantic understanding and classification accuracy are improved, but productivity and scalability deteriorate

Engineering Contradiction:
Improvesemantic understandingVSAvoidprocessing speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system creates an intermediary between automated classification and human expertise. This intermediary presents machine-generated classifications to users for review, allowing human experts to focus only on correcting errors rather than performing all analysis manually, thus maintaining both accuracy and scalability.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The feedback mechanism enables human experts to provide corrections and insights that are automatically incorporated into the machine learning model. This feedback loop allows the system to scale by distributing the workload: machines handle high-speed processing while humans provide targeted feedback to improve semantic understanding.

Inventive Principle:
Principle #23Feedback

4Adaptability or versatility

If the number of classifiers and features is increased to handle diverse data types, then adaptability is improved, but device complexity increases

Engineering Contradiction:
Improvedata type coverageVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system implements a universal classification framework that can handle multiple data types through a single integrated platform. The same core machine learning infrastructure supports various classification tasks across different data formats, reducing the need for separate specialized systems for each data type.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system dynamically adapts its classification capabilities based on the data being processed. Rather than having fixed specialized classifiers for each data type, the system can flexibly configure and adjust its classification models to accommodate diverse data formats, reducing overall system complexity while maintaining versatility.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3742357B1Active featuring in computer-human interactive learning
Publication Date: 2026.04.22 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3742357B1 patent drawingFigure 1
  • EP3742357B1 patent drawingFigure 2
  • EP3742357B1 patent drawingFigure 3

AI summary

A collection of data that is extremely large can be difficult to search and/or analyze. Relevance may be dramatically improved by automatically classifying queries and web pages in useful categories, and using these classification scores as relevance features. A thorough approach may require building a large number of classifiers, corresponding to the various types of information, activities, and products. Creation of classifiers and schematizers is provided on large data sets. Exercising the classifiers and schematizers on hundreds of millions of items may expose value that is inherent to the data by adding usable meta-data. Some aspects include active labeling exploration, automatic regularization and cold start, scaling with the number of items and the number of classifiers, active featuring, and segmentation and schematization.