Speech Translation Confidence via Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional speech recognition systems face inefficiencies in translating spoken utterances from one language to another, often producing multiple possible translations with low confidence, making it difficult to determine the most accurate translation.

Innovation Solution

A speech-processing system that segments spoken sentences into smaller phrases, translates each segment into candidate textual expressions in the first language, and then into candidate textual phrases in a second language, using confidence metrics based on the product of confidence values to identify accurate translations, thereby improving confidence estimation and reducing redundant translations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional speech recognition systems translate utterances using two-stage process (speech-to-text then language translation), then translation functionality is achieved, but multiple possible translations are produced with low confidence making it difficult to determine accurate translation

Engineering Contradiction:

Improvetranslation accuracyVSAvoidtranslation process complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the utterance into multiple phonetic segments and translates each segment independently through multiple candidate paths, then integrates these segment-level translations. This segmentation allows the system to manage complexity by processing smaller units while maintaining overall translation accuracy through integration of segment results with confidence metrics.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a confidence metric dimension to the translation process, transforming the traditional two-stage process into a multi-dimensional approach that incorporates confidence values at each translation stage. This additional dimension enables the system to evaluate and select more accurate translations from multiple candidates, improving reliability without proportionally increasing complexity.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Adaptability or versatility

If the system translates each segment into multiple candidate textual expressions and then into multiple candidate textual phrases, then translation coverage is improved, but the number of possible translations increases making it difficult to identify accurate translations

Engineering Contradiction:

Improvetranslation coverageVSAvoidinformation overload from multiple options

Core Design Contradiction:

Adaptability or versatilityVSLoss of information

Solution Approach 1:

The patent implements feedback mechanisms at multiple levels: confidence values are calculated for each translation candidate based on phonetic segment matches, and these confidence metrics feed back into the selection process. The system uses confidence thresholds and ranking to filter and prioritize candidates, reducing information overload while maintaining comprehensive translation coverage.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent changes parameters such as confidence thresholds and weighting factors to control the number and quality of output translations. By adjusting these parameters, the system can adaptively control the balance between translation coverage and the number of candidates presented, preventing information overload while maintaining versatility.

Inventive Principle:

Principle #35Parameter changes

3Ease of operation

If conventional systems use confidence values to select the most likely translation, then translation selection is simplified, but the confidence values from multiple translation paths are not integrated leading to suboptimal translation accuracy

Engineering Contradiction:

Improvetranslation selection simplicityVSAvoidtranslation accuracy

Core Design Contradiction:

Ease of operationVSReliability

Solution Approach 1:

The patent merges confidence values from multiple translation paths by integrating segment-level confidence metrics with phrase-level confidence metrics. This combining approach allows the system to maintain simplicity in selection while improving accuracy through aggregated confidence information from multiple paths, resolving the contradiction between ease of operation and translation accuracy.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS8914277B1Speech and language translation of an utterance

Publication Date: 2014.12.16 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

According to example configurations, a speech-processing system parses an uttered sentence into segments. The speech-processing system translates each of the segments in the uttered sentence into candidate textual expressions (i.e., phrases of one or more words) in a first language. The uttered sentence can include multiple phrases or candidate textual expressions. Additionally, the speech-processing system translates each of the candidate textual expressions into candidate textual phrases in a second language. Based at least in part on a product of confidence values associated with the candidate textual expressions in the first language and confidence values associated with the candidate textual phrases in the second language, the speech-processing system produces a confidence metric for each of the candidate textual phrases in the second language. The confidence metric can indicate degree to which the candidate textual phrase in the second language is an accurate translation of a respective segment in the utterance.