公开(公告)号：US12045727B2

公开(公告)日：2024-07-23

申请号：US17115464

申请日：2020-12-08

Applicant: NEC Laboratories America, Inc.

Inventor： Renqiang Min ,  Christopher Malon ,  Pengyu Cheng

IPC: G06N3/088 ,  G06F40/20 ,  G06N3/02 ,  G06N3/0442 ,  G06N3/08 ,  G06N3/082 ,  G06N3/086 ,  G10L15/06 ,  G10L15/16 ,  G10L15/22

CPC classification number: G06N3/088 ,  G06F40/20 ,  G06N3/0442 ,  G06N3/08 ,  G06N3/086 ,  G10L15/063 ,  G10L15/16 ,  G10L15/22 ,  G06N3/02 ,  G06N3/082

Abstract: A computer-implemented method is provided for disentangled data generation. The method includes accessing, by a bidirectional Long Short-Term Memory (LSTM) with a multi-head attention mechanism, a dataset including a plurality of pairs each formed from a given one of a plurality of input text structures and given one of a plurality of style labels for the plurality of input text structures. The method further includes training the bidirectional LSTM as an encoder to disentangle a sequential text input into disentangled representations comprising a content embedding and a style embedding based on a subset of the dataset. The method also includes training a unidirectional LSTM as a decoder to generate a next text structure prediction for the sequential text input based on previously generated text structure information and a current word, from a disentangled representation with the content embedding and the style embedding.

公开(公告)号：US20210174213A1

公开(公告)日：2021-06-10

申请号：US17115464

申请日：2020-12-08

Applicant: NEC Laboratories America, Inc.

Inventor： Renqiang Min ,  Christopher Malon ,  Pengyu Cheng

IPC: G06N3/08 ,  G06F40/20

Abstract: A computer-implemented method is provided for disentangled data generation. The method includes accessing, by a bidirectional Long Short-Term Memory (LSTM) with a multi-head attention mechanism, a dataset including a plurality of pairs each formed from a given one of a plurality of input text structures and given one of a plurality of style labels for the plurality of input text structures. The method further includes training the bidirectional LSTM as an encoder to disentangle a sequential text input into disentangled representations comprising a content embedding and a style embedding based on a subset of the dataset. The method also includes training a unidirectional LSTM as a decoder to generate a next text structure prediction for the sequential text input based on previously generated text structure information and a current word, from a disentangled representation with the content embedding and the style embedding.