Semantic Search Explanation via Document Retrieval

Solution Overview

Problem

Machine learning algorithms, particularly those using semantic mappings, produce results that are difficult for humans to interpret due to their complex and opaque nature, making it challenging to understand the reasons behind similarity judgments between complex objects like images, audio clips, and text.

Innovation Solution

A document retrieval system is employed to identify and highlight relevant example objects and features, leveraging embedding mappings and Locality Sensitive Hash (LSH) functions to provide explanations for similarity comparisons by indexing and pre-processing instance objects, allowing users to understand which features contribute to similarity judgments.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If semantic mappings are used to compute similarity between complex objects, then measurement precision is improved, but ease of operation deteriorates because the mappings are opaque and uninterpretable

Engineering Contradiction:

Improvesimilarity measurement precisionVSAvoidinterpretability of similarity judgments

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The patent introduces an intermediary explanation system that mediates between the opaque semantic mapping function and the user. This explanation system generates human-interpretable justifications for similarity judgments by analyzing which features of the input objects contribute most to the computed similarity, thereby making the black-box mapping transparent without altering its precision

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the similarity judgment process into two distinct components: the computational core (semantic mapping) that maintains precision, and the explanation layer that provides interpretability. By separating these functions, the system preserves measurement precision while adding ease of operation through explanations of which specific features drove the similarity assessment

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If embedding mappings are trained on large datasets to optimize accuracy, then measurement precision is improved, but device complexity increases

Engineering Contradiction:

Improvesimilarity measurement accuracyVSAvoidcomplexity of mapping system

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary explanation system that mediates between the opaque semantic mapping function and the user. This explanation system generates human-interpretable justifications for similarity judgments by analyzing which features of the input objects contribute most to the computed similarity, thereby making the black-box mapping transparent without altering its precision

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the similarity judgment process into two distinct components: the computational core (semantic mapping) that maintains precision, and the explanation layer that provides interpretability. By separating these functions, the system preserves measurement precision while adding ease of operation through explanations of which specific features drove the similarity assessment

Inventive Principle:

Principle #1Segmentation

3Measurement precision

If complex machine learning algorithms are used to make decisions, then measurement precision is improved, but ease of operation deteriorates due to opaque inner workings

Engineering Contradiction:

Improvedecision accuracyVSAvoidunderstandability of decisions

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The patent introduces an intermediary explanation system that mediates between the opaque semantic mapping function and the user. This explanation system generates human-interpretable justifications for similarity judgments by analyzing which features of the input objects contribute most to the computed similarity, thereby making the black-box mapping transparent without altering its precision

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent segments the similarity judgment process into two distinct components: the computational core (semantic mapping) that maintains precision, and the explanation layer that provides interpretability. By separating these functions, the system preserves measurement precision while adding ease of operation through explanations of which specific features drove the similarity assessment

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11423319B2Explaining semantic search

Publication Date: 2022.08.23 GALIANO MEDICAL SOLUTIONS INC

AI summary

The invention uses document retrieval to explain to a human user the properties of a query object that are revealed by a machine learning procedure, lending interpretability to the procedure. A query object is compared to reference objects by transforming the query object and reference objects into representative tokens. Reference objects with many tokens in common with the query object are returned as relevant result objects by a document retrieval system. The token representation furthermore admits comparison between features of the query object and matched features of the reference object or between the query object and groups of reference objects having common features, thus emphasising characteristics of the query and reference objects of semantic importance to the user based on the intention of their search. Embodiments include retrieval of 2-dimensional or 3-dimensional images, audio clips, and text.