Document Summarization Using Coherence Filtering for Mobile Screens
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile platforms, such as smartphones, are not ideally suited for presenting long-form news content due to limited screen size, making it difficult for consumers to access and engage with comprehensive news articles effectively.
Innovation Solution
A system and method for summarizing single document articles using a combination of unsupervised Machine Learning techniques and linguistically motivated rules to generate coherent summaries, which are designed to fit the screen size of mobile devices, ensuring that only relevant and coherent sentences are included in the summary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If long-form news content is presented on mobile platforms, then comprehensive information is provided, but screen size limitations make it difficult to access and engage with the content effectively
Solution Approach 1:
The patent segments long-form news articles into multiple short summaries, each fitting within mobile screen constraints. The system divides the original article into discrete summary units that can be consumed incrementally on mobile devices, resolving the contradiction between providing comprehensive information and maintaining ease of access on limited screens.
Solution Approach 2:
The patent transforms the traditional single-dimensional approach (reading full articles) into a multi-dimensional consumption model where users can access summaries at different levels of detail. The system generates multiple summary versions and allows navigation through different summary depths, enabling comprehensive information access adapted to mobile screen constraints.
2Productivity
If summaries are generated using automated techniques, then productivity is improved, but coherence and quality of summaries may deteriorate
Solution Approach 1:
The patent implements feedback mechanisms where user interactions with summaries (such as engagement metrics, completion rates, and preference data) are used to continuously refine and improve summary generation. This feedback loop enables the system to maintain high coherence and quality while operating at automated speeds, as the system learns from user responses to adjust its summarization approach.
Solution Approach 2:
The patent replaces traditional mechanical text extraction methods with machine learning-based semantic understanding systems. This substitution enables automated generation of coherent summaries by using natural language processing and semantic analysis rather than simple keyword matching or sentence extraction, maintaining reliability while achieving high productivity.
3Loss of information
If multiple sentences are selected for summaries, then information completeness is improved, but the risk of including incoherent sentences increases
Solution Approach 1:
The patent introduces coherence-checking algorithms as intermediary components between sentence selection and final summary assembly. These intermediary systems evaluate the logical flow and semantic consistency of selected sentences, acting as a filter that ensures information completeness while maintaining coherence. The intermediary layer prevents incoherent sentence combinations from appearing in final summaries.
Solution Approach 2:
The patent replaces simple sentence-counting or keyword-matching selection mechanisms with machine learning models that understand semantic relationships and contextual coherence. This substitution enables the system to select multiple sentences that collectively provide complete information while maintaining natural flow and logical consistency, as the AI models can assess coherence beyond surface-level features.
Data Source
AI summary
Systems and methods for summarizing online articles for consumption on a user device are disclosed herein. The system extracts the main body of an article's text from the HTML code of an online article. The system may then classify the extracted article into one of several different categories and removes duplicate articles. The system breaks down the article into its component sentences, and each sentence is classified into one of three categories: (1) potential candidate sentences that may be included in the generated summary; (2) weakly rejected sentences that will not be included in the summary but may be used to generate the summary; and (3) strongly rejected sentences that are not included in the summary. Finally, the system applies a document summarizer to generate quickly readable article summaries, for viewing on the user device, using relevant sentences from the article while maintaining the coherence of the article.