Cognitive Data Preparation Engine for Deep Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing enterprise data sets are difficult to utilize for training and testing cognitive systems due to the need for extensive manual annotation, requiring significant human resources and not being set up for cognitive system utilization.

Innovation Solution

A data preparation engine performs statistical analysis to generate statistical distributions of cases and features correlated with the operation for which the cognitive system is to be trained, selecting a diverse subset of entries for annotation and using a template population-based approach to create an annotated training dataset.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If extensive manual annotation is performed on enterprise datasets, then training accuracy for cognitive systems is improved, but human resource requirements and time consumption increase significantly

Engineering Contradiction:

Improvetraining accuracyVSAvoidannotation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary statistical analysis on the enterprise dataset to generate statistical distributions of cases and features before annotation begins. This preliminary action identifies the most informative subsets of data that will yield the highest training accuracy, allowing annotation efforts to be focused on only those critical portions rather than the entire dataset.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

Instead of annotating the complete enterprise dataset, the system applies partial action by selecting and annotating only a statistically determined subset of cases and features. The statistical distribution analysis identifies the minimum necessary portion of data that captures the essential patterns needed for effective model training, reducing overall annotation workload while maintaining training effectiveness.

Inventive Principle:

Principle #16Partial or excessive action

2Productivity

If statistical analysis is performed to select a subset of cases for annotation, then annotation workload is reduced, but the complexity of the data preparation process increases

Engineering Contradiction:

Improveannotation efficiencyVSAvoiddata preparation complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The system introduces statistical distribution analysis as an intermediary step between raw data and annotation. This intermediary process automatically computes statistical characteristics of the enterprise dataset, generating objective criteria for subset selection. This mediator eliminates the need for manual judgment calls about which data to annotate, replacing subjective complexity with automated statistical computation.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The data preparation system performs self-service by automatically conducting statistical analysis and identifying optimal subsets for annotation without requiring manual intervention. The system autonomously computes statistical distributions, determines which cases and features are most informative, and prepares the annotated dataset, reducing the need for human expertise in the data selection process.

Inventive Principle:

Principle #25Self-service

3Reliability

If the entire enterprise dataset is annotated, then comprehensive training coverage is achieved, but the cost and resources required for annotation increase

Engineering Contradiction:

Improvetraining coverageVSAvoidannotation resources

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The system changes the parameter of data selection from exhaustive (all cases) to selective (statistically determined subset). By computing statistical distributions of cases and features, the system identifies which parameters (specific cases and features) contribute most to training reliability. This parameter change allows achieving comparable training coverage with fewer annotated examples by focusing on the most informative instances.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system extracts only the essential and most informative cases and features from the enterprise dataset based on statistical analysis. Rather than annotating everything, the extraction process identifies and isolates the critical subset that captures the dominant patterns and relationships in the data, removing unnecessary annotation of redundant or less informative instances.

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS11475335B2Cognitive data preparation for deep learning model training

Publication Date: 2022.10.18 INTERNATIONAL BUSINESS MACHINE CORPORATION

AI summary

A mechanism is provided in a data processing system for training a computer implemented model. The mechanism determines an operation for which the computer implemented model is to be trained. The mechanism performs a statistical analysis of an enterprise dataset for an enterprise to generate one or more statistical distributions of cases and features correlated with the operation for which the computer implemented model is to be trained. The mechanism selects a subset of cases in the enterprise dataset for annotation based on the one or more statistical distributions of cases and features. The mechanism annotates the selected subset of cases to generate an annotated training dataset. The mechanism trains the computer implemented model, using the annotated training dataset, to perform the operation.