Sequence-to-Sequence Text Processing for Flexible Table Generation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current named entity extraction methods are inefficient due to a fixed table format and the need for predefined entity types, limiting flexibility and processing efficiency.

Innovation Solution

A text processing method using a sequence-to-sequence model that converts source text into a target sequence and then into a flexible target table, eliminating the need for predefined entity types and allowing for a variety of table formats.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If named entity extraction method with predefined entity types is used, then the text processing can be performed, but the table format is fixed and lacks flexibility

Engineering Contradiction:

Improvetable format flexibilityVSAvoidpredefined entity type requirements

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

Instead of predefining entity types and then extracting entities to form tables, the patent inverts the process by directly generating flexible table formats from text using a sequence-to-sequence model, eliminating the constraint of predefined entity types and fixed two-column formats

Inventive Principle:

Principle #13The other way round (Inversion)

Solution Approach 2:

The patent transforms the static, fixed table format into a dynamic, flexible structure where the table schema can adapt to different extraction tasks. The sequence-to-sequence model generates tables with variable columns and structures based on the input text, rather than forcing all outputs into a predetermined format

Inventive Principle:

Principle #15Dynamics

2Productivity

If predefined entity types are required, then entity extraction can be performed, but the text processing process becomes cumbersome and efficiency decreases

Engineering Contradiction:

Improvetext processing efficiencyVSAvoidprocessing process complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent extracts and removes the cumbersome step of predefined entity type selection from the processing pipeline. By using a sequence-to-sequence model that directly maps text to table outputs, the method eliminates the intermediate step of defining entity types, thereby simplifying the process and improving efficiency

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The sequence-to-sequence model performs self-service by automatically determining the appropriate table structure and content based on the input text, without requiring external guidance or predefined schemas. The model adapts to different extraction tasks autonomously, reducing manual intervention and processing complexity

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS20240176955A1Text processing method, model training method, device, and storage medium

Publication Date: 2024.05.30 SHANGHAI SUIXUNTONG ELECTRONICS TECH CO LTD

AI summary

Provided are a text processing method, a model training method, a device, and a storage medium. The text processing method includes: obtaining a source text; inputting the source text into a sequence-to-sequence model, to obtain a target sequence corresponding to the source text; and converting the target sequence into a target table.