Multilingual intent recognition

Patent metadata

Field	Value
Publication number	US-11900922-B2
Application number	US-202017093673-A
Country	US
Kind code	B2
Filing date	Nov 10, 2020
Priority date	Nov 10, 2020
Publication date	Feb 13, 2024
Grant date	Feb 13, 2024

Abstract

Embodiments of the present invention provide computer implemented methods, computer program products and computer systems. For example, embodiments of the present invention can access one or more intents and associated entities from limited amount of speech to text training data in a single language. Embodiments of the present invention can locate speech to text training data in one or more other languages using the accessed one or more intents and associated entities to locate speech to text training data in the one or more other languages different than the single language. Embodiments of the present invention can then train a neural network based on the limited amount of speech to text training data in the single language and the located speech to text training data in the one or more other languages.

First claim

What is claimed is:

1. A computer-implemented method comprising:
accessing one or more intents and associated entities from limited amount of speech to text training data in a single language;
identifying candidate speech to text training data from one or more other languages, each language differing from the single language based on the accessed intents and associated entities;
pooling shared parameters from the identified candidate speech to text training data with the limited amount of speech to text training data in the single language;
training a neural network using the pooled shared parameters from the identified candidate speech to text training data and the limited amount of speech to text training data in the single language;
in response to receiving media containing speech associated with an unidentified language, identifying, in real time, at least one known language and domain that represents a topic discussed in the received media;
identifying language independent intent from the identified language using the pooled shared parameters and adding a respective label to the identified language and language independent intent; and
adapting the trained neural network to specific domains within the same language by replacing language specific parameters of the trained neural network with a new domain and respective language specific layer while maintaining shared layers.

2. The computer-implemented method of claim 1 , further comprising:
training the neural network for natural language processing based on the limited amount of speech to text training data in the single language and the accessed speech to text training data in the one or more other languages.

3. The computer-implemented method of claim 1 , wherein the limited amount of speech to text training data comprises a single language and interpreted intent and associated entities from voice in the single language.

4. The computer-implemented method of claim 1 , wherein the limited amount of speech to text training data comprises a single language drawn from different datasets comprises a dataset collected to model dialog states and a dataset of voice commands.

5. The computer-implemented method of claim 1 , further comprising:
enabling language switching by accessing a common domain and pooling different language datasets having shared commonalities in intents and entities.

6. The computer-implemented method of claim 1 , wherein speech to text training data represents a set of layers of a neural network comprising a set of nodes with each layer connected to other layers in the set of layers via respective weighted connections.

7. The computer-implemented method of claim 1 , further comprising:
learning language independent constructs for each known language by accessing the pooled shared parameters.

8. A computer program product comprising:
one or more computer readable storage media and program instructions stored on the one or more computer readable storage media, the program instructions comprising:
program instructions to access one or more intents and associated entities from limited amount of speech to text training data in a single language;
program instructions to identify candidate speech to text training data from one or more other languages, each language differing from the single language based on the accessed intents and associated entities;
program instructions to pool shared parameters from the identified candidate speech to text training data with the limited amount of speech to text training data in the single language;
program instructions to train a neural network using the pooled shared parameters from the identified candidate speech to text training data and the limited amount of speech to text training data in the single language;
program instructions to, in response to receiving media containing speech associated with an unidentified language, identify, in real time, at least one known language and domain that represents a topic discussed in the received media;
program instructions to identify language independent intent from the identified language using the pooled shared parameters and adding a respective label to the identified language and language independent intent; and
program instructions to adapt the trained neural network to specific domains within the same language by replacing language specific parameters of the trained neural network with a new domain and respective language specific layer while maintaining shared layers.

9. The computer program product of claim 8 , wherein the program instructions stored on the one or more computer readable storage media further comprise:
program instructions to train the neural network for natural language processing based on the limited amount of speech to text training data in the single language and the accessed speech to text training data in the one or more other languages.

10. The computer program product of claim 8 , wherein the limited amount of speech to text training data comprises a single language and interpreted intent and associated entities from voice in the single language.

11. The computer program product of claim 8 , wherein the limited amount of speech to text training data comprises a single language drawn from different datasets comprises a dataset collected to model dialog states and a dataset of voice commands.

12. The computer program product of claim 8 , wherein the program instructions stored on the one or more computer readable storage media further comprise:
program instructions to enable language switching by accessing a common domain and pooling different language datasets having shared commonalities in intents and entities.

13. The computer program product of claim 8 , wherein speech to text training data represents a set of layers of a neural network comprising a set of nodes with each layer connected to other layers in the set of layers via respective weighted connections.

14. A computer system comprising:
one or more computer processors;
one or more computer readable storage media; and
program instructions stored on the one or more computer readable storage media for execution by at least one of the one or more computer processors, the program instructions comprising:
program instructions to access one or more intents and associated entities from limited amount of speech to text training data in a single language;
program instructions to identify candidate speech to text training data from one or more other languages, each language differing from the single language based on the accessed intents and associated entities;
program instructions to pool shared parameters from the identified candidate speech to text training data with the limited amount of speech to text training data in the single language;
program instructions to train a neural network using the pooled shared parameters from the identified candidate speech to text training data and the limited amount of speech to text training data in the single language;
program instructions to, in response to receiving media containing speech associated with an unidentified language, identify, in real time, at least one known language and domain that represents a topic discussed in the received media;
program instructions to identify language independent intent from the identified language using the pooled shared parameters and adding a respective label to the identified language and language independent intent; and
program instructions to adapt the trained neural network to specific domains within the same language by replacing language specific parameters of the trained neural network w

Assignees

• IBM

Inventors

• Thomas Samuel
• Kuo Hong-Kwang
• Audhkhasi Kartik
• Picheny Michael Alan

Classifications

• G06N3/096

  Transfer learning · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/0985

  Hyperparameter optimisation; Meta-learning; Learning-to-learn · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title