Processing

Please wait...

Settings

Settings

Goto Application

1. WO2022161613 - CROSS-LINGUAL APPARATUS AND METHOD

Publication Number WO/2022/161613
Publication Date 04.08.2022
International Application No. PCT/EP2021/052047
International Filing Date 29.01.2021
IPC
G06F 40/30 2020.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
30Semantic analysis
G06F 40/58 2020.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
40Processing or translation of natural language
58Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
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
G06N 3/08 2006.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
G10L 15/26 2006.1
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
26Speech to text systems
CPC
G06F 40/30
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
30Semantic analysis
G06F 40/58
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
40Processing or translation of natural language
58Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
G06N 3/045
G06N 3/088
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
088Non-supervised learning, e.g. competitive learning
G10L 15/1822
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
08Speech classification or search
18using natural language modelling
1822Parsing for meaning understanding
Applicants
  • HUAWEI TECHNOLOGIES CO., LTD. [CN]/[CN]
  • GRITTA, Milan [SK]/[DE] (US)
Inventors
  • GRITTA, Milan
Agents
  • KREUZ, Georg
Priority Data
Publication Language English (en)
Filing Language English (EN)
Designated States
Title
(EN) CROSS-LINGUAL APPARATUS AND METHOD
(FR) APPAREIL ET PROCÉDÉ TRANSLINGUES
Abstract
(EN) Described is an apparatus (500) and method (400) for cross-lingual training between a source language and at least one target language. The method comprises receiving (401) a plurality of input data elements, each of the plurality of input data elements comprising a first linguistic expression (204) in the source language and a second linguistic expression (205) in the target language, the first and the second linguistic expressions having corresponding meaning in their respective languages; and training a neural network model (208) by repeatedly: i. selecting (402) one of the plurality of input data elements; ii. obtaining (403) a first representation of the first linguistic expression of the selected input data element by means of the neural network model; iii. obtaining (404) a second representation of the second linguistic expression of the selected input data element by means of the neural network model; iv. forming (405) a first loss in dependence on the performance of the neural network model on the first linguistic expression; v. forming (406) a second loss indicative of a similarity between the first representation and the second representation; and vi. adapting (407) the neural network model in dependence on the first and second losses. This may improve the performance of models in cross-lingual natural language understanding and classification tasks.
(FR) L'invention concerne un appareil (500) et un procédé (400) d'entraînement translingue entre un langage source et au moins un langage cible. Le procédé comprend la réception (401) d'une pluralité d'éléments de données d'entrée, chacun de la pluralité d'éléments de données d'entrée comprenant une première expression linguistique (204) dans la langue source et une seconde expression linguistique (205) dans la langue cible, les première et seconde expressions linguistiques ayant une signification correspondante dans leurs langues respectives ; et l'entraînement d'un modèle de réseau neuronal (208) en répétant : i. la sélection (402) de l'un de la pluralité d'éléments de données d'entrée ; ii. l'obtention (403) d'une première représentation de la première expression linguistique de l'élément de données d'entrée sélectionné au moyen du modèle de réseau neuronal ; iii. l'obtention (404) d'une seconde représentation de la seconde expression linguistique de l'élément de données d'entrée sélectionné au moyen du modèle de réseau neuronal ; iv. la formation (405) d'une première perte en fonction de la performance du modèle de réseau neuronal sur la première expression linguistique ; v. la formation (406) d'une seconde perte indiquant une similarité entre la première représentation et la seconde représentation ; et vi. l'adaptation (407) du modèle de réseau neuronal en fonction des première et seconde pertes. Cela permet d'améliorer la performance de modèles dans des tâches de compréhension et de classification de langage naturel interlingue.
Latest bibliographic data on file with the International Bureau