Table Content Documents for Accurate Tabular Data Interpretation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for interpreting tabular data, especially those with complex data or structural relationships, suffer from accuracy issues due to the complexity of converting tables into sentences.
Innovation Solution
A method and system that generate table content documents (TCDs) from tabular data, allowing for the creation of query features based on questions, identifying relevant TCDs, and determining candidate answers by matching query features with TCD features above a threshold, without constructing new sentences or introducing new matter.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If table narration converts tabular data into sentence form, then the data can be processed by natural language models, but the accuracy decreases when tables contain complex data, significant text, or numerous structural relationships
Solution Approach 1:
The patent divides the tabular data into multiple independent table content documents (TCDs), where each TCD represents a specific cell or data element with its row header, column header, and value. This segmentation allows the system to process complex tables by handling small, manageable units rather than attempting to convert the entire complex table structure into a single sentence, thereby maintaining accuracy while enabling NLP processing.
Solution Approach 2:
The patent introduces table content documents (TCDs) as an intermediary representation between the original tabular data and natural language queries. Instead of directly converting tables to sentences (which loses precision), the TCDs serve as a structured intermediate format that preserves all table relationships and can be efficiently queried and processed by NLP models.
2Loss of information
If the system processes complex tabular data with numerous structural relationships, then comprehensive information can be retrieved, but the computational complexity and processing time increase
Solution Approach 1:
By segmenting the table into individual TCDs, the system can process and search through data elements independently rather than analyzing the entire complex table structure at once. This allows comprehensive information retrieval while reducing processing time through parallel processing of individual TCDs.
Solution Approach 2:
The system pre-processes the tabular data into TCD format with extracted features (row headers, column headers, values, and relationships) before receiving queries. This preliminary organization of data structures enables faster query processing and reduces computational complexity during actual data retrieval operations.
Data Source
AI summary
One or more table content documents (TCDs) can be constructed for a set of tabular data or portion thereof. A set of query features corresponding to a question can be matched to one or more TCDs. A respective candidate answer can be generated for each of the one or more TCDs having a set of features matching the set of query features above a threshold. Zero or more candidate answers can be output to a user consumable data object.