Processing

Please wait...

Settings

Settings

Goto Application

1. WO2021239631 - NEURAL MACHINE TRANSLATION METHOD, NEURAL MACHINE TRANSLATION SYSTEM, LEARNING METHOD, LEARNING SYSTEM, AND PROGRAMM

Publication Number WO/2021/239631
Publication Date 02.12.2021
International Application No. PCT/EP2021/063697
International Filing Date 21.05.2021
IPC
G06F 40/44 2020.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
40Processing or translation of natural language
42Data-driven translation
44Statistical methods, e.g. probability models
G06F 40/242 2020.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
20Natural language analysis
237Lexical tools
242Dictionaries
G06N 3/02 2006.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
CPC
G06F 40/242
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
20Natural language analysis
237Lexical tools
242Dictionaries
G06F 40/44
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
40Processing or translation of natural language
42Data-driven translation
44Statistical methods, e.g. probability models
G06N 3/0445
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0445Feedback networks, e.g. hopfield nets, associative networks
G06N 3/0454
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0454using a combination of multiple neural nets
G06N 3/084
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
084Back-propagation
Applicants
  • IP.APPIFY GMBH [DE]/[DE]
  • EISSFELLER, Thomas [DE]/[DE]
Inventors
  • EISSFELLER, Thomas
Agents
  • WESER & KOLLEGEN PATENTANWÄLTE PARTMBB
Priority Data
10 2020 114 046.026.05.2020DE
Publication Language English (en)
Filing Language English (EN)
Designated States
Title
(EN) NEURAL MACHINE TRANSLATION METHOD, NEURAL MACHINE TRANSLATION SYSTEM, LEARNING METHOD, LEARNING SYSTEM, AND PROGRAMM
(FR) PROCÉDÉ DE TRADUCTION AUTOMATIQUE NEURONALE, SYSTÈME DE TRADUCTION AUTOMATIQUE NEURONALE, PROCÉDÉ D'APPRENTISSAGE, SYSTÈME D'APPRENTISSAGE ET PROGRAMME
Abstract
(EN) A neural machine translation method for translating a source text in a source language into a target text in a target language is provided which obtains the source text, a source auxiliary entry, such as a source dictionary entry, in the source language and a target auxiliary entry, such as a target dictionary entry, in the target language as explicit inputs and computes, using a multi-layer encoder-decoder network, target token information (215) regarding a target token of the target text, by combining information (213) computed based on at least one previously computed or initial target token (203) with information (211, 212, 214) computed based on the source text, the source auxiliary entry, and the target auxiliary entry. A corresponding learning method is provided which performs a learning step using a record including a source text and a target text from which a source phrase and a target phrase are extracted using a computed alignment, wherein the source and the target phrase are used as the source dictionary entry and the target dictionary entry for computing target token information (215) according to the neural machine translation method and a loss computed by comparing the target token information(215) with information corresponding to the actual target text is used to adjust at least some of the parameters of the multi-layer encoder-decoder network.
(FR) L'invention concerne un procédé de traduction automatique neuronale pour traduire un texte source d'une langue source en un texte cible d'une langue cible, qui permet d'obtenir le texte source, une entrée auxiliaire source, telle qu'une entrée de dictionnaire source, dans la langue source, et une entrée auxiliaire cible, telle qu'une entrée de dictionnaire cible, dans la langue cible en tant qu'entrées explicites, et de calculer, à l'aide d'un réseau codeur-décodeur multicouche, des informations de jeton cible (215) concernant un jeton cible du texte cible, en combinant des informations (213) calculées sur la base d'au moins un jeton cible initial ou précédemment calculé (203) avec des informations (211, 212, 214) calculées sur la base du texte source, de l'entrée auxiliaire source et de l'entrée auxiliaire cible. L'invention concerne un procédé d'apprentissage correspondant qui effectue une étape d'apprentissage à l'aide d'un enregistrement comprenant un texte source et un texte cible à partir desquels une phrase source et une phrase cible sont extraites à l'aide d'un alignement calculé, la phrase source et la phrase cible étant utilisées en tant qu'entrée de dictionnaire source et entrée de dictionnaire cible pour calculer des informations de jeton cible (215) selon le procédé de traduction automatique neuronale, et une perte calculée par comparaison des informations de jeton cible (215) avec des informations correspondant au texte cible réel est utilisée pour ajuster au moins certains des paramètres du réseau codeur-décodeur multicouche.
Related patent documents
Latest bibliographic data on file with the International Bureau