Phonological clustering

Patent metadata

Field	Value
Publication number	US-10943580-B2
Application number	US-201815977507-A
Country	US
Kind code	B2
Filing date	May 11, 2018
Priority date	May 11, 2018
Publication date	Mar 9, 2021
Grant date	Mar 9, 2021

Abstract

Methods and systems for phonological clustering are disclosed. A method includes: segmenting, by a computing device, a sentence into a plurality of tokens; determining, by the computing device, a plurality of phoneme variants corresponding to the plurality of tokens; clustering, by the computing device, the plurality of phoneme variants; creating, by the computing device, an initial vectorization of the plurality of phoneme variants based on the clustering; embedding, by the computing device, the initial vectorization of the plurality of phoneme variants into a deep learning model; and determining, by the computing device, a radial set of phoneme variants using the deep learning model.

First claim

What is claimed is:

1. A method comprising:
segmenting, by a computing device, a sentence into a plurality of tokens;
determining, by the computing device, a plurality of phoneme variants corresponding to the plurality of tokens;
clustering, by the computing device, the plurality of phoneme variants;
creating, by the computing device, an initial vectorization of the plurality of phoneme variants based on the clustering;
embedding, by the computing device, the initial vectorization of the plurality of phoneme variants into a deep learning model; and
determining, by the computing device, a radial set of phoneme variants using the deep learning model.

2. The method according to claim 1 , further comprising using the radial set of phoneme variants in a text-to-speech process.

3. The method according to claim 2 , wherein the text-to-speech process uses particular phoneme variants from the radial set based upon geographic or demographic information associated with a target audience.

4. The method according to claim 1 , wherein the creating the initial vectorization comprises assigning contiguous numbers or other identification to each phoneme variant in a cluster.

5. The method according to claim 1 , wherein the deep learning model uses a feedforward architecture for a neural network.

6. The method according to claim 1 , wherein a repository of audio and corresponding text is input into the deep learning model as training data.

7. The method according to claim 6 , wherein the training data further comprises metadata including geographic or demographic data associated with the audio.

8. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computing device to cause the computing device to:
cluster a plurality of phoneme variants based upon associated phonemes;
create an initial vectorization of the plurality of phoneme variants based on the clustering;
embed the initial vectorization of the plurality of phoneme variants into a deep learning model; and
determine a radial set of phoneme variants using the deep learning model.

9. The computer program product according to claim 8 , wherein the plurality of phoneme variants is determined by segmenting a sentence into a plurality of tokens and determining the plurality of phoneme variants corresponding to the plurality of tokens,
wherein a first radial set of the phoneme variants are variants of a first phoneme of the phonemes,
a second radial set of the phoneme variants are variants of a second phoneme of the phonemes,
the first phoneme and the second phoneme are different phonemes, and
at least one of the phoneme variants belongs to both the first radial set and the second radial set.

10. The computer program product according to claim 8 , the program instructions further executable by the computing device to cause the computing device to use the radial set of phoneme variants in a text-to-speech process.

11. The computer program product according to claim 10 , wherein the text-to-speech process uses particular phoneme variants from the radial set based upon geographic or demographic information associated with a target audience.

12. The computer program product according to claim 8 , wherein the creating the initial vectorization comprises assigning contiguous numbers or other identification to each phoneme variant in a cluster.

13. The computer program product according to claim 8 , wherein the deep learning model uses a feedforward architecture for a neural network.

14. The computer program product according to claim 8 , wherein a repository of audio and corresponding text is input into the deep learning model as training data.

15. The computer program product according to claim 14 , wherein the training data further comprises metadata including geographic or demographic data associated with the audio.

16. A system comprising:
a hardware processor, a computer readable memory, and a computer readable storage medium associated with a computing device;
program instructions configured to segment a sentence into a plurality of tokens;
program instructions configured to determine a plurality of phoneme variants corresponding to the plurality of tokens;
program instructions configured to cluster the plurality of phoneme variants;
program instructions configured to create an initial vectorization of the plurality of phoneme variants based on the clustering;
program instructions configured to embed the initial vectorization of the plurality of phoneme variants into a deep learning model; and
program instructions configured to determine a plurality of radial sets of phoneme variants using the deep learning model, each of the plurality of radial sets comprising a distinct phoneme,
wherein the program instructions are stored on the computer readable storage medium for execution by the hardware processor via the computer readable memory.

17. The system according to claim 16 , further comprising program instructions configured to use the radial set of phoneme variants in a text-to-speech process.

18. The system according to claim 17 , wherein the text-to-speech process is configured to use particular phoneme variants from the radial set based upon geographic or demographic information associated with a target audience.

19. The system according to claim 16 , wherein the deep learning model uses a feedforward architecture for a neural network.

20. The system according to claim 16 , wherein a repository of audio and corresponding text is input into the deep learning model as training data.

Assignees

• IBM

Inventors

• Trim Craig M
• Ganci Jr John M
• Bostick James E
• Fonseca Carlos A

Classifications

• G10L13/00 Primary

  Speech synthesis; Text to speech systems · CPC title

• G10L15/063 Primary

  Training · CPC title

• G06F18/2321

  using statistics or function optimisation, e.g. modelling of probability density functions · CPC title

• G06F18/23

  Clustering techniques · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title