Similarity Rating Histogram for Electronic Discovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Reviewers face inefficiency in finding similar documents during electronic discovery due to the manual and time-consuming process of setting similarity thresholds, as current methods require trial and error to achieve a reasonable number of similar documents.
Innovation Solution
A system that determines counts of similar documents for a set of similarity ratings based on co-occurring terms and presents a graphical user interface (GUI) with a visual representation, such as a histogram, allowing users to input a user-specified similarity rating threshold to efficiently retrieve relevant documents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If reviewers manually set similarity rating thresholds and run searches to determine the number of similar documents, then they can find documents that meet their criteria, but the process becomes inefficient and time-consuming
Solution Approach 1:
The system pre-calculates and stores the counts of similar documents for multiple similarity ratings before the reviewer needs them. When a reviewer selects a reference document, the system immediately retrieves the pre-computed similarity counts from storage rather than calculating them on-demand, eliminating the time-consuming trial-and-error process of setting thresholds.
Solution Approach 2:
The system introduces an intermediary data structure (storage mechanism) that holds pre-computed similarity counts between the reference document and candidate documents. This intermediary layer allows reviewers to quickly query similarity information without triggering expensive real-time calculations, acting as a buffer that decouples the query operation from the computation operation.
2Measurement precision
If reviewers use a high similarity threshold to ensure document quality, then the number of similar documents decreases, but it may become too few to be useful
Solution Approach 1:
The system dynamically adjusts the similarity threshold based on reviewer preferences and document characteristics. Rather than using a fixed high threshold, the system allows reviewers to explore multiple threshold levels and automatically adapts the threshold to balance precision and quantity, ensuring that the retrieved document set is both high-quality and sufficiently large.
Solution Approach 2:
The system changes the similarity rating parameter to explore different thresholds. By pre-computing similarity counts across a range of ratings and presenting them to reviewers, the system enables dynamic parameter adjustment without re-running searches, allowing reviewers to find the optimal balance between similarity precision and document quantity.
3Quantity of substance
If reviewers use a low similarity threshold to increase the number of similar documents, then more documents are retrieved, but the quality and relevance may decrease
Solution Approach 1:
The system provides feedback to reviewers by displaying the distribution of similarity ratings and the count of documents at each threshold level. This feedback mechanism allows reviewers to understand the impact of different threshold choices on both quantity and quality, enabling them to make informed decisions about the appropriate similarity threshold for their review task.
4Speed
If the system pre-calculates similarity counts for multiple ratings, then reviewers can quickly access similarity information, but the initial processing time and computational resources increase
Solution Approach 1:
The system performs preliminary calculation of similarity counts during document indexing or batch processing, storing these pre-computed values in a data structure for rapid retrieval. This shifts the computational burden from interactive query time to background processing time, making the difference between immediate response and delayed response.
Solution Approach 2:
The system calculates similarity counts for more similarity ratings than any single reviewer will likely need (excessive action). By pre-computing a comprehensive set of similarity counts across the full range of possible ratings, the system ensures that any future query can be answered immediately without additional calculation, accepting the extra initial computation as a worthwhile investment.
Data Source
AI summary
A count of documents similar to a reference document is determined based on a plurality of similarity ratings. Each similarity rating may be based on a number of co-occurring terms between the reference document and the corresponding similar documents. A graphical user interface (GUI) may be provided. The GUI may include a GUI element that is associated with the similar documents. Furthermore, the GUI element may include a visual representation of a number of documents for each similarity rating that are retrievable based on a selection of the corresponding similarity rating. The GUI element may be provided prior to retrieving one of the similar documents.