Table Query Answering With Cell Relevance Scoring for LLM Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine-learning models, particularly large language models (LLMs), struggle with accurately answering questions from tables due to their susceptibility to noise and inability to process the structural composition of tables effectively, leading to inaccurate results and inefficient resource usage.
Innovation Solution
A system that calculates relevance scores for table cells based on a query, using a combination of unsupervised and weakly-supervised methods to focus the LLMs on relevant content, suppressing irrelevant information and enhancing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional machine-learning models process entire table content, then comprehensive search coverage is achieved, but accuracy decreases due to noise and irrelevant content
Solution Approach 1:
The patent segments the table into individual cells and calculates relevance scores for each cell based on the query. This segmentation allows the system to identify and focus on relevant cells while suppressing irrelevant ones, thereby improving search result accuracy without processing the entire table content uniformly.
Solution Approach 2:
The patent applies local quality by assigning different relevance scores to different cells within the table. Cells with higher relevance scores receive more attention from the machine-learning model, while cells with lower scores are suppressed. This differential treatment based on local relevance improves overall search accuracy.
2Reliability
If conventional techniques process all table content, then complete information is considered, but computational resource efficiency deteriorates
Solution Approach 1:
The patent extracts only the relevant cells from the table based on calculated relevance scores. By taking out and processing only these relevant cells rather than the entire table, the system maintains search result accuracy while significantly reducing computational resource consumption.
Solution Approach 2:
The patent applies partial action by processing only a subset of table cells that have high relevance scores, rather than processing all cells. This partial processing approach achieves the necessary search accuracy without the excessive computational resources required for complete table processing.
3Reliability
If relevance scoring is implemented for table cells, then search accuracy improves, but system complexity increases
Solution Approach 1:
The patent introduces relevance scores as an intermediary mechanism between the query and the table cells. These scores act as a mediator that guides the machine-learning model's attention, improving search accuracy without requiring complex modifications to the underlying model architecture.
Data Source
AI summary
Content relevance based table query answering is described. In one or more examples, a query and a table are received. The table includes a plurality of cells. A plurality of scores for calculated that correspond to the plurality of cells based on the query. One or more machine-learning models are then leveraged to generate a search result from the query, table, and scores, which is presented in a user interface for display.