Contextual N-Gram Tagging With LDA for Dynamic Content Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing content tagging systems rely on manual tagging and limited training data, leading to inaccurate and inefficient categorization, especially when adapting to dynamic content changes.

Innovation Solution

A system that utilizes a latent Dirichlet allocation (LDA) model to generate data-dependent content tags, supplemented by deep learning models, to enhance automated content categorization and reduce reliance on manual tagging, by generating contextual N-grams and using Bayesian inference for probabilistic clustering.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual tagging is used for content categorization, then tag accuracy can be maintained through human judgment, but productivity is reduced due to the time-consuming nature of manual processes

Engineering Contradiction:
Improvetag accuracyVSAvoidcontent processing speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system performs preliminary automated tagging using LDA models to generate candidate tags before final selection, preparing the content in advance for more accurate and efficient categorization without requiring complete manual tagging from scratch

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary automated tagging system that bridges manual tagging accuracy and automated processing speed, using LDA-generated tags as intermediate results that can be reviewed or directly applied based on confidence thresholds

Inventive Principle:
Principle #24Intermediary (Mediator)

2Device complexity

If limited training data is used for content tagging models, then the system complexity is reduced, but measurement precision deteriorates due to insufficient learning samples

Engineering Contradiction:
Improvemodel training requirementsVSAvoidcategorization accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The LDA model performs self-service by automatically generating topic tags from the content itself without requiring external labeled training data, enabling the system to create its own training signals and improve accuracy while maintaining low complexity

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system changes the parameter of training data requirements by shifting from supervised learning (requiring labeled data) to unsupervised topic modeling (requiring only raw text), thereby maintaining accuracy while reducing data requirements

Inventive Principle:
Principle #35Parameter changes

3Ease of operation

If generic tagging methods are used for content categorization, then ease of operation is improved through standardized processes, but adaptability deteriorates when facing dynamic content changes

Engineering Contradiction:
Improvetagging process simplicityVSAvoidresponse to content changes
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The system implements dynamic adaptability by continuously applying LDA topic modeling to new content, allowing the tagging system to automatically adjust to changing content patterns while maintaining a simple automated operation process

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The LDA-based tagging system serves multiple functions simultaneously: it categorizes content, discovers topics, adapts to new content types, and provides interpretable tags, thereby achieving both ease of operation and adaptability through a single unified approach

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4111375B1Automated content tagging with latent dirichlet allocation of contextual word embeddings
Publication Date: 2025.11.26 PALO ALTO NETWORKS INC
  • EP4111375B1 patent drawingFigure 1
  • EP4111375B1 patent drawingFigure 2
  • EP4111375B1 patent drawingFigure 3

AI summary

Dynamic content tags are generated as content is received by a dynamic content tagging system. A natural language processor (NLP) tokenizes the content and extracts contextual N-grams based on local or global context for the tokens in each document in the content. The contextual N-grams are used as input to a generative model that computes a weighted vector of likelihood values that each contextual N-gram corresponds to one of a set of unlabeled topics. A tag is generated for each unlabeled topic comprising the contextual N-gram having a highest likelihood to correspond to that unlabeled topic. Topic-based deep learning models having tag predictions below a threshold confidence level are retrained using the generated tags, and the retrained topic-based deep learning models dynamically tag the content.