ASR Heterograph Resolution via Knowledge Graph Context

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional automatic speech recognition (ASR) systems face challenges in accurately processing heterographs, where different words with similar phonetic sounds lead to incorrect outputs or user frustration, as they often randomly select one word or require user clarification.

Innovation Solution

The system encodes verbal input into phoneme sequences and uses a knowledge graph to determine the context of the first word, selecting the word from a heterograph set that is most closely related based on the context and relationships in the graph, thereby resolving ambiguity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional ASR systems randomly select one word from heterograph sets, then the system operates quickly and simply, but the output accuracy deteriorates due to incorrect word selection

Engineering Contradiction:
Improveoutput accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary component (heterograph resolution module with knowledge graph) that mediates between the phoneme recognition stage and the final text output. This intermediary analyzes contextual relationships to disambiguate heterographs, improving accuracy without requiring complete system redesign.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system performs preliminary actions by pre-building a knowledge graph that stores contextual relationships between words before ASR processing occurs. During speech recognition, this pre-computed knowledge structure enables rapid disambiguation of heterographs based on contextual clues from previously recognized words.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If conventional ASR systems request user clarification for heterographs, then output accuracy improves, but the user experience deteriorates due to frustration and additional input requirements

Engineering Contradiction:
Improveoutput accuracyVSAvoiduser experience
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The ASR system performs self-service by automatically resolving heterograph ambiguity using contextual analysis and knowledge graph querying. The system independently determines the correct word without requiring user intervention, maintaining high accuracy while preserving ease of operation.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If the system uses contextual analysis and knowledge graphs to resolve heterographs, then output accuracy improves, but processing time increases

Engineering Contradiction:
Improveoutput accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The knowledge graph is pre-computed and stored before ASR processing, containing contextual relationships between words. During speech recognition, the system only needs to query this pre-built structure using previously recognized words as keys, rather than performing comprehensive contextual analysis in real-time, thus minimizing processing time overhead.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system performs partial contextual analysis by utilizing only the necessary portion of the knowledge graph relevant to the current speech input and previously recognized words. Rather than analyzing all possible contextual relationships, the system focuses on disambiguating specific heterographs using minimal required context, reducing processing time while maintaining accuracy.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3364408B1Systems and methods for performing ASR in the presence of heterographs
Publication Date: 2021.05.19 ADEIA GUIDES INC
  • EP3364408B1 patent drawingFigure 1~2
  • EP3364408B1 patent drawingFigure 3~4
  • EP3364408B1 patent drawingFigure 5

AI summary

Systems and methods for performing ASR in the presence of heterographs are provided. Verbal input is received from the user that includes a plurality of utterances. A first of the plurality of utterances is matched to a first word. It is determined that a second utterance in the plurality of utterances matches a plurality of words that is in a same heterograph set. It is identified which one of the plurality of words is associated with a context of the first word. A function is performed based on the first word and the identified one of the plurality of words.