Abstractive-Extractive Summarization Graph and Vector Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing text summarization methods struggle to generate concise, natural-language summaries of multiple documents, often relying on time-consuming manual processes and producing generic or unnatural summaries.

Innovation Solution

The use of machine learning techniques, specifically abstractive summarization with part-of-speech tagging and graph data structures, combined with an extractive approach employing numerical vector representations, to generate ranked candidate summary sentences and select natural-language summaries that reflect the content of multiple documents.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If manual template-based summarization is used, then summaries can be domain-specific, but the process is time-consuming and requires manual effort for each domain

Engineering Contradiction:

Improveease of summary generationVSAvoidtime for template creation

Core Design Contradiction:

Ease of manufactureVSLoss of time

Solution Approach 1:

The system automatically generates summaries without requiring manual template creation. The machine learning model processes documents and generates summaries autonomously, eliminating the need for human experts to manually design templates for each domain.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

Manual template-based summarization is replaced with an automated machine learning system. The mechanical process of manually creating and maintaining templates is substituted with an automated computational approach that uses trained models to generate summaries.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If frequency-based or semantics-based keyword extraction is used, then characteristic terms can be identified, but actual summary sentences are not generated

Engineering Contradiction:

Improvekeyword extraction accuracyVSAvoidsummary generation capability

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system uses keyword extraction as an intermediary step to identify important terms, then feeds these into a sequence-to-sequence model that generates complete summary sentences. The keyword extraction serves as a bridge between raw text and final summaries, enabling both precise term identification and full sentence generation.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The summarization process is divided into distinct stages: keyword extraction, sequence-to-sequence modeling, and summary generation. This segmentation allows each component to specialize in a specific task, improving overall system effectiveness.

Inventive Principle:

Principle #1Segmentation

3Productivity

If abstractive summarization is used, then concise summaries can be generated, but the summaries may appear generic or robotic

Engineering Contradiction:

Improvesummary generation speedVSAvoidnatural language quality

Core Design Contradiction:

ProductivityVSEase of operation

Solution Approach 1:

The system applies different processing qualities to different parts of the summarization task. Keyword extraction uses frequency and semantic analysis for precision, while the sequence-to-sequence model uses attention mechanisms to maintain natural language flow in the generated summaries, combining structured analysis with natural language generation.

Inventive Principle:

Principle #3Local quality

4Area of stationary object

If summaries are reduced for small format displays, then information fits the display, but readability and completeness may be compromised

Engineering Contradiction:

Improvedisplay area utilizationVSAvoidinformation completeness

Core Design Contradiction:

Area of stationary objectVSLoss of information

Solution Approach 1:

The system extracts only the most salient information from documents to create condensed summaries. By using attention mechanisms and relevance scoring, it identifies and extracts key information that maintains completeness while reducing overall volume for small format displays.

Inventive Principle:

Principle #2Taking out (Extraction)

Data Source

PatentUS12032905B2Methods and systems for summarization of multiple documents using a machine learning approach

Publication Date: 2024.07.09 AMADEUS SAS

AI summary

An abstractive technique and an extractive technique are used to generate concise natural-language summaries of related text documents. The abstractive step generates a machine summary by constructing a graph with nodes representing unique pairs of tokens and corresponding parts-of-speech (POS), and with edge sequences representing token/POS pairs comprising sentences of a corresponding topic group from the text documents. Ranked candidate summary sentences are generated using subgraphs of the graph having initial and final nodes corresponding with valid sentence start and end pairs. The machine summary includes representative summary sentence(s) selected from each topic group's ranked candidates. The extractive step generates a natural-language summary from the machine summary by computing, for each topic group, numerical suitability measures providing comparisons between the representative summary sentence and sentences of the topic group. The natural-language summary is composed by selecting, for each topic group, a preferred summary sentence based on the numerical suitability measures.