Author-Style Model for Neural NLG
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current natural language generation (NLG) applications struggle to produce stylistically appealing content for publications that require specific authorship, as they primarily focus on data-driven materials and are unable to meet the demand for instantaneous, cost-effective generation of diverse content across various communication channels.
Innovation Solution
A method utilizing a neural network-based system that learns the style of an author by analyzing their original works, allowing for automated generation of publications that mimic the author's style, incorporating an author-style model, audience model, and genre model to ensure stylistic consistency and audience appeal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing NLG applications focus on generating data-driven materials with predetermined formats, then generation efficiency is improved, but stylistic quality and authorship characteristics deteriorate
Solution Approach 1:
The system segments the NLG process into distinct modules: a style model that captures authorship characteristics separately from the content generation process, allowing stylistic quality to be optimized independently while maintaining generation efficiency through automated processing
Solution Approach 2:
A style model acts as an intermediary between the content generation system and the final output, transferring authorship characteristics to generated texts without requiring human authors to be directly involved in each generation process
2Manufacturing precision
If human authors produce publications across multiple communication channels, then stylistic quality is maintained, but production cost and time consumption increase
Solution Approach 1:
The system creates a computational copy of an author's writing style by analyzing their original works and capturing stylistic patterns, enabling automated generation of publications that mimic the author's voice without requiring the author's direct involvement, thereby reducing production costs while maintaining stylistic quality
Solution Approach 2:
The style model is built in advance by analyzing an author's existing works, preparing the system to generate stylistically consistent publications on demand across multiple channels without requiring real-time human author intervention
3Speed
If the demand for instantaneous publication across various channels increases, then market responsiveness is improved, but resource requirements and production cost worsen
Solution Approach 1:
The system enables self-service publication generation by automating the content creation process with AI models that can independently produce stylistically consistent publications across multiple channels without requiring human authors for each piece, reducing resource requirements while maintaining market responsiveness
Data Source
AI summary
Methods, computer program products, and systems are presented. The methods include, for instance: obtaining a style feed including a plurality of original works by an author. An author-style model for the author is built based on the style feed by use of a selected neural network, and a publication is generated in the style of the author based on the author-style model.