Machine Reading Comprehension for Large-Document Query Answering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine reading comprehension systems struggle to efficiently process and analyze large documents, such as those with tens to hundreds of pages, due to the significant time and computational resources required for training and processing.
Innovation Solution
A machine reading comprehension system that divides documents into sections, processes each section separately using multiple processing circuitries, including neural networks and information retrieval systems, to determine scores indicating relevance to a query, and ranks sections based on these scores to predict an answer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If a machine reading comprehension system processes large documents (tens to hundreds of pages), then the system can analyze comprehensive content, but the processing time and computational resources required increase significantly
Solution Approach 1:
The patent divides a large document into multiple sections, allowing the system to process each section separately and independently. This segmentation enables the system to handle large documents by breaking them down into manageable units, reducing the computational burden and processing time while maintaining the ability to analyze comprehensive content.
Solution Approach 2:
The system performs preliminary processing by dividing the document into sections before the main analysis. This preliminary action of sectioning the document allows for more efficient subsequent processing, as each section can be analyzed independently and in parallel, significantly reducing the overall processing time for large documents.
2Reliability
If a machine reading comprehension system processes large documents, then the system can provide comprehensive answers, but substantial processing capabilities and considerable time are required
Solution Approach 1:
By segmenting the document into sections, the system can apply multiple processing circuitries to different sections independently. This approach maintains answer accuracy by ensuring comprehensive analysis while reducing the complexity burden on any single processing unit, as each circuitry handles only a portion of the overall task.
Solution Approach 2:
The patent combines multiple processing circuitries to work on different sections of the document simultaneously. This merging of processing capabilities allows the system to maintain high answer accuracy through comprehensive analysis while distributing the computational load, thereby reducing the complexity requirement for any individual processing unit.
3Adaptability or versatility
If training data and processing power are increased to handle large documents, then the system's analytical capability improves, but the time and resources required for training and processing increase
Solution Approach 1:
The system segments the document processing task across multiple processing circuitries, allowing each circuitry to handle a specific section with dedicated analytical capabilities. This segmentation enables the system to maintain high adaptability and versatility while distributing computational resources efficiently, reducing the overall energy and resource consumption compared to a single-unit approach.
Solution Approach 2:
The patent applies partial processing by having multiple processing circuitries handle different sections of the document rather than one circuitry processing the entire document. This partial action approach allows the system to achieve comprehensive analytical capability while using computational resources more efficiently, as each circuitry performs a subset of the total processing task.
Data Source
AI summary
A machine reading comprehension system (MRCS) can analyze a larger-sized document that includes multiple pages to predict an answer to a query. For example, the document can have two, five, tens, or hundreds of pages. The MRCS divides the document into multiple sections with each section including a portion of the document. Each section is processed separately by one or more processing circuitries to determine a score for that section. The score indicates how related the section is to the query and/or a probability that the section provides a possible answer to the query. Once all of the sections have been analyzed, the sections are ranked by their scores and a subset of the ranked sections are processed again to determine a predicted answer to the query.