公开(公告)号：EP4394652A1

公开(公告)日：2024-07-03

申请号：EP23219472.0

申请日：2023-12-21

申请人： Pocketalk Corporation

发明人： KAWATAKE, Hajime

IPC分类号： G06F40/47 ,  G06F40/51 ,  G06F40/58 ,  G10L15/00

CPC分类号： G06F40/51 ,  G06F40/58 ,  G06F40/47

摘要： A translation engine evaluation system (1) includes a parallel translation list display control unit (120) that displays a list of parallel translations, in which the pre-translation text and the post-translation text are associated with each other, on the terminal (100), a parallel translation list selection receiving unit (122) that receives a selection of a parallel translation from the plurality of parallel translations included in the list, a comparative translation unit (218) that causes an alternative translation engine to translate the pre-translation text included in the selected parallel translation into the second language to generate a comparative post-translation text, a comparative translation list display control unit (130) that displays the comparative post-translation text on the terminal (100), and an individual evaluation information obtaining unit (210) that obtains individual evaluation information indicating evaluation of a user of the terminal (100) on the comparative post-translation text.

公开(公告)号：EP4388462A2

公开(公告)日：2024-06-26

申请号：EP22801377.7

申请日：2022-10-06

申请人： DeepMind Technologies Limited

发明人： SAVINOV, Nikolay ,  CHUNG, Junyoung ,  BINKOWSKI, Mikolaj ,  VAN DEN OORD, Aaron Gerard Antonius ,  ELSEN, Erich Konrad

IPC分类号： G06N3/088 ,  G06N3/0455 ,  G06N3/047 ,  G06N3/084

CPC分类号： G06N3/0455 ,  G06N3/088 ,  G06N3/084 ,  G06N3/047 ,  G06F40/47 ,  G10L15/16 ,  G06V10/82 ,  G06V20/70 ,  G10L15/063

公开(公告)号：EP4369246A1

公开(公告)日：2024-05-15

申请号：EP23206525.0

申请日：2023-10-27

申请人： SDL Limited

发明人： De Vrieze, Erik ,  Sârbe, Viorel Daniel ,  Gallo, Giovanni

IPC分类号： G06F40/47 ,  G06F40/51 ,  G06N3/02

CPC分类号： G06F40/47 ,  G06F40/51 ,  G06N3/02

摘要： A computer implemented method (100) of and system for evaluating translations using a trained neural network. The inputs to the neural network for training (110) and evaluating (120) translations include a source content, translated content, source and target language identifier. Additional inputs for neural network training and evaluating translations can include translation origin before adaptation identifiers, file types, translator identifiers, customer identifiers, and content domain identifiers. The neural network outputs (140) an indication that the translation is either correct or incorrect. Corrections by a reviewer or translator can be used in feedback for further training of the evaluation neural network.