Speech Keyword Classification for Ad Bidding Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing speech recognition systems face challenges in accurately transcribing keywords with non-standard pronunciations, leading to incorrect or irrelevant advertisements being displayed, which results in wasted resources and potential loss of revenue for advertisers.

Innovation Solution

A keyword bidding system that classifies candidate adwords based on their recognition by the speech recognizer, providing feedback to advertisers on appropriate terms and suggesting similar keywords to reach the intended audience, using a combination of vocabulary analysis, text-to-speech systems, language models, and acoustic models to assign weights and suggest alternative keywords.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If speech recognition is used to transcribe user queries, then input convenience is improved, but transcription accuracy deteriorates for keywords with nonstandard pronunciations

Engineering Contradiction:
Improveinput convenienceVSAvoidtranscription accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The system performs preliminary analysis of candidate keywords before final transcription, checking whether they exist in the speech recognizer's vocabulary and comparing expected pronunciations with actual transcriptions to identify and correct errors proactively

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system provides feedback to advertisers about which of their keywords are recognizable by the speech system and suggests alternative pronunciations, enabling continuous improvement of keyword recognition accuracy

Inventive Principle:
Principle #23Feedback

2Measurement precision

If advertisers bid on keywords with nonstandard pronunciations, then ad targeting precision is improved, but ad relevance deteriorates due to incorrect transcription

Engineering Contradiction:
Improvead targeting precisionVSAvoidad relevance
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The system pre-validates advertiser keywords against the speech recognizer's vocabulary and identifies potential transcription errors before campaigns launch, allowing advertisers to adjust their keyword selections

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system introduces an intermediary validation layer between keyword selection and speech recognition, acting as a bridge that translates between advertiser intent and speech system capabilities

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If the speech recognizer uses standard vocabulary only, then transcription accuracy is improved, but keyword coverage deteriorates

Engineering Contradiction:
Improvetranscription accuracyVSAvoidkeyword coverage
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system segments keyword handling into two categories: standard vocabulary keywords that work directly with speech recognition, and nonstandard keywords that require special validation and alternative pronunciation strategies

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically adjusts its approach based on keyword characteristics, using different validation and transcription strategies depending on whether a keyword exists in the standard vocabulary or requires special handling

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3848927B1Speech recognition for keywords
Publication Date: 2024.11.06 GOOGLE LLC
  • EP3848927B1 patent drawingFigure 1
  • EP3848927B1 patent drawingFigure 2
  • EP3848927B1 patent drawingFigure 3

AI summary

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speech recognition are disclosed. In one aspect, a method includes receiving a candidate adword from an advertiser. The method further includes generating a score for the candidate adword based on a likelihood of a speech recognizer generating, based on an utterance of the candidate adword, a transcription that includes a word that is associated with an expected pronunciation of the candidate adword. The method further includes classifying, based at least on the score, the candidate adword as an appropriate adword for use in a bidding process for advertisements that are selected based on a transcription of a speech query or as not an appropriate adword for use in the bidding process for advertisements that are selected based on the transcription of the speech query.