Search Query Expansion for Content Safety

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Search engines face challenges in blocking unwanted content, such as adult material, due to misclassifications of search queries, especially with short queries lacking explicit terms or ambiguous terms, leading to leakage in search results and potential public relations issues for companies.

Innovation Solution

The method involves expanding original search queries by adding neutral words with equal co-occurrence statistics in both positive and negative classes of a domain, using pointwise mutual information, to provide additional context for classification models, thereby improving query classification accuracy without requiring additional data or retraining models.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If search engines use classification models to block unwanted content, then content safety is improved, but misclassification of search queries (especially short or ambiguous queries) leads to unwanted content leakage

Engineering Contradiction:

Improvecontent safetyVSAvoidquery classification accuracy

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The system performs preliminary query expansion by adding neutral words to search queries before classification. This preliminary action enriches the query representation with additional context that helps the classification model make more accurate decisions, especially for short or ambiguous queries that would otherwise be misclassified.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

Neutral words act as intermediaries between the original search query and the classification model. These neutral words provide bridging context that connects the query to relevant content categories without introducing bias, enabling more accurate classification while maintaining content safety.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If search engines block unwanted content strictly, then content safety is improved, but legitimate content may be blocked causing false positives

Engineering Contradiction:

Improvecontent safetyVSAvoidfalse positive blocking

Core Design Contradiction:

ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The system changes the parameter of query representation by expanding queries with neutral words. This parameter change enriches the input to the classification model, allowing it to distinguish more accurately between legitimate and unwanted content, thereby reducing false positives while maintaining safety.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system uses feedback from the classification model's analysis of expanded queries to adjust content distribution. By analyzing the expanded query representation, the model can feedback with more accurate classifications that reduce both false negatives (unwanted content leakage) and false positives (legitimate content blocking).

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11934468B2Content distribution control

Publication Date: 2024.03.19 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

The present disclosure relates to methods and systems for controlling content distribution. The methods and systems receive an original search query and expand the original search query by adding one or more neutral words to create an expanded search query. The methods and systems automatically classify the original search query using the expanded search query in an accurate manner and use the classification to control content distribution.