Notification System Using Text Embeddings for Real-Time Content Filtering

Solution Overview

Problem

Existing notification systems struggle to process the large volume of data from third-party content in a timely manner, often resulting in delayed delivery of relevant information, which can put individuals at an information disadvantage, especially in fields like finance where real-time insights are crucial.

Innovation Solution

A system that utilizes a notification application, text embedding service, and machine learning models to identify and filter relevant content based on context and keyword topics, enabling real-time or near real-time notifications by converting text into embeddings for analysis and comparison against reference content, thereby prioritizing content that matches specified criteria.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If existing notification systems process large volume of third-party content using keyword-based filtering, then they can identify relevant content, but they are unable to process the data in a timely manner resulting in delayed notifications

Engineering Contradiction:

Improvecontent relevance identificationVSAvoiddata processing speed

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces traditional mechanical keyword-matching systems with machine learning models that use natural language processing and semantic analysis. These models convert text to embeddings and perform vector-based similarity comparison, enabling both high-speed processing of large volumes of content and accurate identification of relevant content based on meaning rather than just keyword presence.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the fundamental parameters of content analysis by transforming text from discrete keyword tokens into continuous vector embeddings. This parameter transformation allows the system to process content at scale while capturing semantic relationships, thereby improving both processing speed and relevance identification accuracy simultaneously.

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If existing systems use comprehensive keyword matching to ensure accurate content identification, then they can identify relevant content, but they generate false positives and miss contextually relevant content

Engineering Contradiction:

Improvecontent relevance accuracyVSAvoidcontextual meaning

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The patent replaces mechanical keyword-matching systems with machine learning models that use natural language processing and semantic analysis. These models convert text to embeddings and perform vector-based similarity comparison, enabling both high-speed processing of large volumes of content and accurate identification of relevant content based on meaning rather than just keyword presence.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the fundamental parameters of content analysis by transforming text from discrete keyword tokens into continuous vector embeddings. This parameter transformation allows the system to process content at scale while capturing semantic relationships, thereby improving both processing speed and relevance identification accuracy simultaneously.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If notification systems process and analyze all third-party content in detail, then they can identify relevant content accurately, but they cannot provide real-time or near real-time notifications

Engineering Contradiction:

Improvecontent analysis accuracyVSAvoidnotification delay

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-processing and converting content into embedding representations in advance, storing these embeddings for rapid comparison. When a notification is requested, the system can quickly compare new content against the pre-processed embeddings without performing full analysis, enabling real-time notifications with high accuracy.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces traditional mechanical keyword-matching systems with machine learning models that use natural language processing and semantic analysis. These models convert text to embeddings and perform vector-based similarity comparison, enabling both high-speed processing of large volumes of content and accurate identification of relevant content based on meaning rather than just keyword presence.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11775357B1Relevant notifications of third party content

Publication Date: 2023.10.03 AMERICAN EXPRESS TRAVEL RELATED SERVICES CO INC

AI summary

Disclosed are various embodiments for generating relevant notifications of content generated by third party data sources. In some embodiments, a system comprises a computing device and machine readable instructions. The computing device includes a processor and a memory. The machine-readable instructions can be stored in the memory that, when executed by the processor, cause the computing device to receive content from a third party data source based at least in part on a keyword topic. An embedding for the content is generated. The system can classify a portion of the content as associated with the keyword topic. Organizations can be identified from the portions of the content. The system can generate a list from the organizations identified in the content and transmit a notification to a client device regarding the content.