ML Submission Summarization for Bias Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for sorting and ranking textual documents, such as in idea contests or RFPs, are time-consuming and prone to human bias, especially when dealing with large volumes of submissions, and existing computer processing is hindered by lengthy documents.
Innovation Solution
A machine learning-based method that preprocesses submissions by converting sentences into tokens, generating summaries by ranking sentences based on topic relevance, creating n-grams, randomizing them, and determining best fit lines in multidimensional space, then analyzes for uniqueness and predefined technology presence, scoring submissions accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If human analysis is used to evaluate submissions, then selection accuracy can be maintained, but processing time increases significantly and human bias occurs
Solution Approach 1:
The patent introduces an automated text processing system as an intermediary between the large volume of submissions and the final selection decision. This system pre-processes, summarizes, and ranks submissions using algorithms, serving as a mediator that reduces the workload for human evaluators while maintaining selection quality.
Solution Approach 2:
The evaluation process is segmented into multiple stages: automated preprocessing, summary generation, initial filtering, and final human review. This segmentation allows different types of analysis to be performed at appropriate stages, reducing the time burden on humans while preserving accuracy where needed.
2Measurement precision
If all submissions are processed in detail, then selection accuracy improves, but processing complexity increases
Solution Approach 1:
The system applies different levels of processing quality to different submissions based on their characteristics. High-priority or ambiguous submissions receive more detailed analysis, while clear-cut cases receive streamlined processing. This local differentiation optimizes the balance between precision and complexity.
Solution Approach 2:
The system performs partial processing on all submissions (summary generation and key feature extraction) and then applies excessive/detailed analysis only where necessary. This approach ensures baseline evaluation precision while avoiding unnecessary complexity for cases that don't require it.
3Productivity
If automated computer processing is used, then processing speed increases, but handling of lengthy documents becomes more time-consuming
Solution Approach 1:
The system extracts and processes only the essential information from lengthy documents - generating summaries, identifying key features, and extracting relevant entities. This extraction approach maintains high processing speed while effectively handling long documents by focusing computational resources on the most important content.
Solution Approach 2:
The system performs preliminary processing actions on all submissions before detailed evaluation - including text normalization, summary generation, and initial filtering. This preliminary action reduces the average document length and complexity that subsequent processing stages must handle, thereby maintaining productivity.
Data Source
AI summary
The disclosed system and method can improve a selection process, such as selecting winners in a contest or request for proposal, by narrowing the submissions to a reasonable number of top submissions that can be more meaningfully considered by a decision making body. By using machine learning during the selection process, human biases may be avoided, processing speed is significantly increased, accuracy of review and scoring is improved, and auditability of results is possible. The disclosed system and method improve the process of selecting one or more submissions by summarizing the submissions into a minimal number of words that can be substituted for the full text of the submission in downstream processes of the selection process. Fewer words used in the downstream processes can improve the speed and efficiency of the downstream processes.