Document Processing Engine for Personalized News Feeds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current online news aggregation systems often fail to provide users with relevant articles due to keyword search limitations, leading to under or over-inclusive results, where terms like 'feline' might not retrieve articles about 'cats' and a nutritionist might receive irrelevant articles about software or hardware.
Innovation Solution
A document processing engine that generates a list of entities and a document vector for each article, associating textual representations with concepts from a taxonomy, allowing for personalized news feeds by linking relevant articles based on user interests and automatically inserting hyperlinks to enhance user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If keyword search is used for article retrieval, then the system is simple to operate, but the retrieval accuracy is poor leading to under or over-inclusive results
Solution Approach 1:
The patent introduces an intermediary layer between the user's keyword search and the article database. This intermediary includes a concept hierarchy structure and relevance scoring mechanism that translates simple keywords into sophisticated concept-based matching, thereby improving retrieval accuracy without requiring the user to interact with the complex system directly
Solution Approach 2:
The system performs preliminary actions by pre-building a concept hierarchy from the article corpus and pre-computing relevance metrics. This allows the system to quickly retrieve accurate results when users perform keyword searches, as the heavy computational work has already been done in advance
2Measurement precision
If keyword search is used for article retrieval, then the system is fast to execute, but the retrieval precision is poor causing irrelevant articles to be returned
Solution Approach 1:
The system performs preliminary action by pre-processing the article corpus to build a concept hierarchy and pre-computing relevance metrics for articles against concepts. This allows query processing to be fast while still achieving high precision, as the computational heavy lifting has been done in advance
Solution Approach 2:
The patent applies local quality by computing detailed relevance scores only for articles that match the query concepts at a threshold level. Instead of uniformly processing all articles with the same level of detail, the system focuses computational resources on locally relevant articles, improving precision without proportionally increasing processing time
Data Source
AI summary
Selectively delivering an article is disclosed. A user preference is received. A document is also received. An entity pair comprising a concept included in a concept taxonomy and a textual representation included in the document is determined. The document is categorized based at least in part on the concept. A user associated with the user preference is selectively notified of the document.