Conceptual Search Highlighting for Document Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current document search techniques lack precision in identifying relevant features within documents, requiring manual review of large volumes of search results, even when machine learning is applied, leading to inefficiencies and wasted time.
Innovation Solution
A computer-implemented method and system that performs conceptual searches by receiving input concepts, calculating concept scores for content items by dividing them into subsections, determining similarity metrics, and displaying relevant subsections with visual indicators, utilizing embedding machine learning models to generate embedded vectors for improved search accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual review of search results is performed to ensure accuracy, then search result precision is improved, but time consumption increases significantly
Solution Approach 1:
The patent segments documents into smaller units (sentences, phrases, or spans) and applies highlighting to specific segments rather than entire documents. This segmentation allows the system to process and evaluate smaller units independently, reducing the overall time required to analyze large volumes of search results while maintaining precision through targeted evaluation of relevant segments.
Solution Approach 2:
The patent replaces manual mechanical review with an automated computational system that uses machine learning models to generate embeddings and calculate similarity metrics. This substitution eliminates the need for human reviewers to manually examine each document, dramatically reducing time consumption while maintaining or improving precision through consistent algorithmic evaluation.
2Productivity
If general conceptual search is applied to reduce document volume, then the number of documents to review is reduced, but the search results fail to provide insight regarding relevant features
Solution Approach 1:
The patent applies local quality by providing different levels of detail and highlighting to different parts of the search results. Specifically, it highlights specific segments (sentences, phrases, or spans) within documents that are most relevant to the search query, rather than treating all documents uniformly. This allows reviewers to quickly identify and focus on the most important features and insights within reduced document sets.
Solution Approach 2:
The patent performs preliminary action by pre-processing documents to generate embeddings and pre-identifying relevant segments before the actual search and review process. The system calculates similarity metrics and generates highlights in advance, so that when reviewers examine the reduced set of search results, the relevant features are already identified and visually distinguished, eliminating the need for secondary analysis.
3Reliability
If reviewers must locate important portions of documents manually, then in-depth analysis can be performed, but additional wasted time is spent on unimportant tasks
Solution Approach 1:
The patent extracts and isolates the important portions of documents by generating visual highlights on specific segments (sentences, phrases, or spans) that are most relevant to the search query. This extraction process separates the valuable information from the rest of the document content, allowing reviewers to immediately focus on the extracted important portions without spending time locating or evaluating unimportant sections.
Solution Approach 2:
The patent uses color changes (visual highlighting) to distinguish important portions of documents from non-important portions. By applying different visual indicators such as background colors, text coloring, or other visual cues to highlighted segments, the system enables reviewers to quickly identify and prioritize relevant content, eliminating time spent on unimportant tasks while maintaining the ability to perform in-depth analysis on the highlighted portions.
Data Source
AI summary
Systems and methods for performing a conceptual search and highlighting relevant portions of a content item are provided. The methods may include: receiving one or more input concepts from a user device; calculating concept scores for the one or more input concepts for one or more content items; based on the concept scores for the one or more content items, retrieving a set of content items; identifying one or more concept subsections of the content items in the set of content items, wherein a concept subsection is a subsection of a content item that corresponds to a particular input concept of the one or more input concepts; and causing the user device to display a particular content item in the set of content items, wherein the display of the particular content item includes visual indicators associated with the identified one or more concept subsections for the particular content item.


