Affective Summarization Neural Network Architecture

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional text summarization systems fail to account for affective preferences of users, leading to summaries that may irritate or confuse users and result in poor comprehension and engagement.

Innovation Solution

An affective summarization system utilizing a combination of a summarization neural network and an affect predictor neural network to generate text summaries based on user-specific affective preferences, determining an affect level and embedding sequence to produce summaries that align with user-defined levels of formality, informality, politeness, and other psycho-linguistic preferences.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional summarization systems generate text summaries without considering affective preferences, then the summarization process is simple and fast, but the user comprehension and engagement are poor

Engineering Contradiction:

Improveuser comprehensionVSAvoidsummarization system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the summarization neural network and affect predictor neural network into a unified affective summarization system. The affect predictor network determines affect levels from the input text, and this affect information is integrated into the summarization process through embeddings generation, enabling the system to produce summaries that are both informative and affectively appropriate.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The affect predictor neural network performs preliminary analysis of the input text to determine affect levels before the summarization process begins. This affect information is then embedded into the summarization workflow, allowing the system to proactively adjust summary generation to match user affective preferences rather than reacting after summary creation.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If conventional summarization systems use standard neural networks, then the system structure is simple, but the summaries do not account for psychological or linguistic preferences of users

Engineering Contradiction:

Improveaffective preference adaptationVSAvoidneural network architecture complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The affective summarization system is designed to handle multiple affective dimensions (formality, politeness, emotionality) through a unified architecture. The embeddings generator creates affect-aware representations that can be applied across different summarization contexts, making the system adaptable to various user preferences without requiring separate specialized systems for each affective dimension.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The affect embeddings serve as an intermediary between the affect predictor network and the summarization decoder. These embeddings translate affect level predictions into a format that can be seamlessly integrated into the existing summarization workflow, bridging the gap between affect analysis and summary generation without requiring complete architectural overhaul.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Reliability

If affective information is incorporated during summary generation, then user engagement improves, but the processing time and computational resources increase

Engineering Contradiction:

Improveuser engagementVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system segments the affective summarization process into distinct modular components: affect prediction, embeddings generation, encoder processing, and decoder generation. This segmentation allows each component to be optimized independently and enables parallel processing where possible, reducing overall processing time while maintaining affective awareness throughout the summarization pipeline.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS10891427B2Machine learning techniques for generating document summaries targeted to affective tone

Publication Date: 2021.01.12 ADOBE INC

AI summary

An affective summarization system provides affective text summaries directed towards affective preferences of a user, such as psychological or linguistic preferences. The affective summarization system includes a summarization neural network and an affect predictor neural network. The affect predictor neural network is trained to provide a target affect level based on a word sequence, such as a word sequence for an article or other text document. The summarization neural network is trained to provide a summary sequence based on the target affect level and on the word sequence for the text document.