Exception-Limited Phonetic Decision Tree Training

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional text-to-speech systems face inefficiencies due to large phonetic decision trees containing extraneous branches for handling phonetic exceptions, leading to increased processing time and storage requirements, with no automated process to eliminate these exceptions.

Innovation Solution

A semi-automated method for training an exception-limited phonetic decision tree, where phonetic exceptions are identified and stored in an exception dictionary, allowing for the creation of a phonetic tree with minimal branches for exceptional words, using a training engine and phonetic tree generation engine to categorize and remove incorrect phonetizations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a large quantity of phonetically exceptional words are included in the training data, then the phonetic tree can handle more exceptional cases, but the processing time and storage space increase

Engineering Contradiction:

Improveability to handle phonetic exceptionsVSAvoidprocessing time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The phonetic tree is segmented into two distinct components: a streamlined phonetic tree for standard words and a separate exception dictionary for phonetically exceptional words. This segmentation allows the system to process most words efficiently through the simplified tree while handling exceptions separately, thus reducing overall processing time while maintaining the ability to handle exceptional cases.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Phonetically exceptional words are extracted from the main training data set and stored in a separate exception dictionary. This extraction removes extraneous branches from the phonetic tree, creating a more efficient structure for standard words, while the exception dictionary preserves the ability to handle exceptional cases without bloating the main tree structure.

Inventive Principle:

Principle #2Taking out (Extraction)

2Adaptability or versatility

If a large quantity of phonetically exceptional words are included in the training data, then the phonetic tree can handle more exceptional cases, but the storage space required increases

Engineering Contradiction:

Improveability to handle phonetic exceptionsVSAvoidstorage space

Core Design Contradiction:

Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The phonetic data structure is segmented into a compact phonetic tree for standard words and a separate exception dictionary. This segmentation reduces the storage requirements of the main phonetic tree by removing exceptional cases, while the exception dictionary stores only the necessary exception data in a space-efficient manner.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Exceptional words are extracted from the main phonetic tree structure and stored separately in an exception dictionary. This extraction eliminates redundant storage of exception handling logic within the main tree, reducing overall storage requirements while preserving exception handling capability.

Inventive Principle:

Principle #2Taking out (Extraction)

3Device complexity

If manual processes are used to eliminate phonetic exceptions from training data, then the phonetic tree can be streamlined, but the process becomes prohibitively time-consuming

Engineering Contradiction:

Improvephonetic tree complexityVSAvoidtraining time

Core Design Contradiction:

Device complexityVSLoss of time

Solution Approach 1:

The system employs an automated process that enables the phonetic tree to self-optimize by automatically identifying phonetically exceptional words, separating them into an exception dictionary, and constructing a streamlined phonetic tree for standard words. This self-service automation eliminates the need for manual intervention while achieving tree simplification, thus reducing both complexity and training time.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The automated process changes the parameter of phonetic exception handling from manual curation to algorithmic identification and separation. By transforming the training process into an automated procedure that systematically identifies and extracts exceptional words, the system achieves tree simplification without the prohibitive time cost of manual processes.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS8027834B2Technique for training a phonetic decision tree with limited phonetic exceptional terms

Publication Date: 2011.09.27 CERENCE OPERATING CO

AI summary

The present invention discloses a method for training an exception-limited phonetic decision tree. An initial subset of data can be selected and used for creating an initial phonetic decision tree. Additional terms can then be incorporated into the subset. The enlarged subset can be used to evaluate the phonetic decision tree with the results being categorized as either correctly or incorrectly phonetized. An exception-limited phonetic tree can be generated from the set of correctly phonetized terms. If the termination conditions for the method have been determined to be unsatisfactorily met, then steps of the method can be repeated.