Adaptive language translation using context features

Patent metadata

Field	Value
Publication number	US-11947925-B2
Application number	US-202016879886-A
Country	US
Kind code	B2
Filing date	May 21, 2020
Priority date	May 21, 2020
Publication date	Apr 2, 2024
Grant date	Apr 2, 2024

Abstract

A user input in a source language is received. A set of contextual data is received. The user input is encoded into a user input feature vector. The set of contextual data is encoded into a context feature vector. The user input feature vector and the context feature vector are used to generate a fusion vector. An adaptive neural network is trained to identify a second context feature vector, based on the fusion vector. A second user input in the source language is received for translation into a target language. The adaptive neural network is used to determine, based on the second context feature vector, a second user input feature vector. The second user input feature vector is decoded, based on the source language and the target language, into a target language output. A user is notified of the target language output.

First claim

What is claimed is:

1. A method for providing language translations, the method comprising:
encoding, from a user input in a source language, a user input feature vector comprising a set of user input features, wherein the user input is associated with a first user;
determining a set of contextual data for the user input, wherein determining the set of contextual data for the user input comprises determining a dialect based on a spoken accent of the user input;
encoding, from the set of contextual data for the user input, a context feature vector;
generating a fusion vector from the user input feature vector and the context feature vector, by using a neural network;
generating, in real-time, a target language output based on the fusion vector and a target language by using a long short-term memory as a recurrent neural network, wherein the LSTM is trained to identify the context feature vector from the fusion vector, and wherein the target language output is generated based on the context feature vector; and
notifying a second user of the target language output.

2. The method of claim 1 , further comprising notifying the first user of the target language output.

3. The method of claim 1 , wherein the set of contextual data includes a set of users, a set of demographic data for each user within the set of users, a dialect of the target language, and a dialect of the source language.

4. The method of claim 3 , wherein the set of demographic data for each user includes an age, a linguistic gender, an education level, a set of cultural information, a familiarity level, and a familial relationship.

5. The method of claim 4 , wherein the set of demographic data for each user includes a language formality level.

6. The method of claim 1 , wherein using the adaptive neural network to generate the target language output includes performing Long Short Term Memory techniques on the fusion vector.

7. The method of claim 6 , and wherein training the adaptive neural network to identify the context feature vector includes adjusting a weight or a bias of the adaptive neural network.

8. The method of claim 1 , wherein software is provided as a service in a cloud environment to perform the method.

9. A computer program product for providing language translations, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a device to cause the device to:
encode, from a user input in a source language, a user input feature vector comprising a set of user input features, wherein the user input is associated with a first user;
determine a set of contextual data for the user input, wherein determining the set of contextual data for the user input comprises determining a dialect based on a spoken accent of the user input;
encode, from the set of contextual data for the user input, a context feature vector;
generate a fusion vector from the user input feature vector and the context feature vector, by using a neural network;
generate, in real-time, a target language output based on the fusion vector and a target language by using a long short-term memory as a recurrent neural network, wherein the LSTM is trained to identify the context feature vector from the fusion vector, and wherein the target language output is generated based on the context feature vector; and
notify a second user of the target language output.

10. The computer program product of claim 9 , wherein the program instructions further cause the device to notify the first user of the target language output.

11. The computer program product of claim 9 , wherein the set of contextual data includes a set of users, a set of demographic data for each user within the set of users, a dialect of the target language, and a dialect of the source language.

12. The computer program product of claim 11 , wherein the set of demographic data for each user includes an age, a linguistic gender, an education level, a set of cultural information, a familiarity level, and a familial relationship.

13. The computer program product of claim 12 , wherein the demographic data for each user includes a language formality level.

14. The computer program product of claim 9 , wherein using the adaptive neural network to generate the target language output includes performing Long Short Term Memory techniques on the fusion vector.

15. A system for providing language translations, comprising:
a memory with program instructions included thereon; and
a processor in communication with the memory, wherein the program instructions cause the processor to:
encode, from a user input in a source language, a user input feature vector comprising a set of user input features, wherein the user input is associated with a first user;
determine a set of contextual data for the user input, wherein determining the set of contextual data for the user input comprises determining a dialect based on a spoken accent of the user input;
encode, from the set of contextual data for the user input, a context feature vector;
generate a fusion vector from the user input feature vector and the context feature vector by using a neural network;

generate, in real-time, a target language output based on the fusion vector and a target language by using a long short-term memory as a recurrent neural network, wherein the LSTM is trained to identify the context feature vector from the fusion vector, and wherein the target language output is generated based on the context feature vector; and
notify a second user of the target language output.

16. The system of claim 15 , wherein the program instructions further cause the processor to notify the first user of the target language output.

17. The system of claim 15 , wherein the set of contextual data includes a set of users, a set of demographic data for each user within the set of users, a dialect of the target language, and a dialect of the source language.

18. The system of claim 17 , wherein the set of demographic data for each user includes an age, a linguistic gender, an education level, a set of cultural information, a familiarity level, and a familial relationship.

19. The system of claim 18 , wherein the demographic data for each user includes a language formality level.

20. The system of claim 15 , wherein using the adaptive neural network to generate the target language output includes performing Long Short Term Memory techniques on the fusion vector.

Assignees

• IBM

Inventors

• Mei Lei
• Yin kun yan
• Hu Yan
• Ruan Qi
• Han yan feng

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0455

  Auto-encoder networks; Encoder-decoder networks · CPC title

• G06N3/0442

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

• G06F40/58 Primary

  Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title