Machine translation method and apparatus

• US 10,380,264 B2
• Filed: 02/10/2017
• Issued: 08/13/2019
• Est. Priority Date: 08/16/2016
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Lee, Hodong
• Primary Examiner(s): Riley, Marcus T
• Application Number:
  US15/429,709
• Publication Number:
  US 20180052829A1
• Time in Patent Office: 914 Days
• CPC Class Codes: G06F 40/268 Morphological analysis
• CPC Class Codes: G06F 40/42 Data-driven translation
• CPC Class Codes: G06F 40/53 Processing of non-Latin text kana-to-kanji conversion G06F40/129; vowelisation G06F40/232
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/048 Activation functions
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A machine translation method and apparatus are provided. The machine translation apparatus generates a feature vector of a source sentence from the source sentence, where the source sentence being is written in a first language, and converts the generated feature vector of the source sentence to a feature vector of a normalized sentence. The machine translation apparatus generates a target sentence from the feature vector of the normalized sentence, wherein the target sentence corresponding corresponds to the source sentence and being is written in a second language.

Claims

1. A processor-implemented method of performing machine translation, the method comprising:

receiving, at an interface device, a source sentence written in a first language;

generating, by a neural network and at a processor, a feature vector of the source sentence by inputting the source sentence to input layers of the neural network;

converting, at the processor and using the neural network, the feature vector of the source sentence to a feature vector of a normalized sentence, the normalized sentence being generated from omitting any one or any combination of any two or more of a vocabulary, a morpheme, or a symbol from the source sentence without replacing the vocabulary, the morpheme, or the symbol;

generating, at the processor and using the neural network, a target sentence from the feature vector of the normalized sentence, wherein the target sentence corresponds to the source sentence and is written in a second language; and

providing, at the processor, a result of the generating of the target sentence from output layers of the neural network.

2. The machine translation method of claim 1, wherein the feature vector of the normalized sentence is a closest feature vector to the generated feature vector of the source sentence, among pre-determined feature vectors of normalized sentences.

3. The machine translation method of claim 1, wherein the normalized sentence is a sentence generated by a substitution of any one or any combination of two or more of a morpheme, a vocabulary, a word, or a phrase included in the source sentence.

4. The machine translation method of claim 1, wherein the normalized sentence is a sentence generated by changing word spacing in the source sentence.

5. The machine translation method of claim 1, wherein the normalized sentence is a sentence generated by changing a word order in the source sentence.

6. The machine translation method of claim 1, wherein the source sentence is a sentence generated by recognizing a voice signal in the first language received from a user.

10. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the method of claim 1.

18. The machine translation method of claim 1, wherein the normalized sentence is a sentence generated by changing a word order to a first order, a second order subsequent to the first order, and a third order subsequent to the second order, wherein the first order comprises a subject as a first word, the second order comprises an object as a first word, and the third order comprises an adverb as a first word.

19. The machine translation method of claim 1, wherein the normalized sentence is generated based on taxonomy mnemonics, morphological structure, and Bayesian probability theory.

20. The machine translation method of claim 1, wherein the normalized sentence is generated based on postpositions.

7. A processor-implemented method of performing machine translation, the method comprising:

receiving, at an interface device, a source sentence written in a first language;

generating, by a neural network and at a processor, a feature vector of the source sentence by inputting the source sentence to input layers of the neural network;

converting, at the processor and using the neural network, the feature vector of the source sentence to a feature vector of a target sentence, wherein the target sentence is written in a second language; and

generating, at the processor and using the neural network, a normalized sentence from the feature vector of the target sentence by transforming the target sentence, by omitting any one or any combination of any two or more of a vocabulary, a morpheme, or a symbol from the source sentence without replacing the vocabulary, the morpheme, or the symbol.

8. The machine translation method of claim 7, wherein the generating of the normalized sentence comprises:

selecting a closest feature vector to the feature vector of the target sentence from feature vectors of pre-determined normalized sentences; and

generating the normalized sentence from the selected closest feature vector.

9. The machine translation method of claim 7, wherein the normalized sentence is a sentence generated by substituting a vocabulary and a morpheme in the target sentence, or changing word spacing and a word order in the target sentence.

11. A machine translation apparatus, comprising:

an interface device configured to receive a source sentence written in a first language;

a memory comprising a neural network;

a processor configured to;

based on a communication with the memory via a bus, generate, by the neural network, a feature vector of the source sentence by inputting the source sentence to input layers of the neural network;

convert, using the neural network, the generated feature vector of the source sentence to a feature vector of a normalized sentence;

generate, using the neural network, a target sentence from the feature vector of the normalized sentence, the normalized sentence being generated from omitting any one or any combination of any two or more of a vocabulary, a morpheme, or a symbol from the source sentence without replacing the vocabulary, the morpheme, or the symbol, wherein the target sentence corresponding to the source sentence is written in a second language; and

provide a result of the generating of the target sentence from output layers of the neural network.

12. The machine translation apparatus of claim 11, wherein the feature vector of the normalized sentence is a closest feature vector to the generated feature vector of the source sentence among feature vectors of pre-determined normalized sentences.

13. The machine translation apparatus of claim 11, wherein the normalized sentence is a sentence generated by a substitution of any one or any combination of any two or more of a morpheme, a vocabulary, a word, or a phrase included in the source sentence.

14. The machine translation apparatus of claim 11, wherein the normalized sentence is a sentence generated by changing word spacing in the source sentence.

15. The machine translation apparatus of claim 11, wherein the normalized sentence is a sentence generated by changing a word order in the source sentence.

16. The machine translation apparatus of claim 11, wherein the source sentence is a sentence generated by recognizing a voice signal in the first language received from a user.

17. The machine translation apparatus of claim 11, wherein the neural network comprises an encoder and a decoder, the decoder comprises a normalization layer that converts the target sentence to a normalized sentence and an attention layer that allows vocabularies in the target sentence to be mapped by a one-to-many relationship or a many-to-one relationship.