公开(公告)号：US20250022459A1

公开(公告)日：2025-01-16

申请号：US18220910

申请日：2023-07-12

Applicant: Adobe Inc.

Inventor： Viet Dac Lai ,  Trung Bui ,  Seunghyun Yoon ,  Quan Tran ,  Hao Tan ,  Hanieh Deilamsalehy ,  Abel Salinas ,  Franck Dernoncourt

IPC: G10L15/183 ,  G10L15/065

Abstract: The disclosed method generates helpful training data for a language model, for example, a model implementing a punctuation restoration task, for real-world ASR texts. The method uses a reinforcement learning method using a generative AI model to generate additional data to train the language model. The method allows the generative AI model to learn from real-world ASR text to generate more effective training examples based on gradient feedback from the language model.

公开(公告)号：US20250111129A1

公开(公告)日：2025-04-03

申请号：US18478221

申请日：2023-09-29

Applicant: ADOBE INC.

Inventor： Viet Dac Lai ,  Franck Dernoncourt

IPC: G06F40/14 ,  G06F40/279

Abstract: A method, apparatus, and non-transitory computer readable medium for natural language processing are described. Embodiments of the present disclosure include obtaining a document comprising a first event mention and a second event mention. Some embodiments generate a dependency tree based on the document. The dependency tree is pruned by removing an irrelevant word to obtain a pruned dependency tree. Subevent relation information is generated for the first event mention and the second event mention based on the pruned dependency tree.

公开(公告)号：US20240330669A1

公开(公告)日：2024-10-03

申请号：US18116129

申请日：2023-03-01

Applicant: ADOBE INC.

Inventor： Amir Pouran Ben Veyseh ,  Viet Dac Lai ,  Franck Dernoncourt

IPC: G06N3/08 ,  G06N3/0475

CPC classification number: G06N3/08 ,  G06N3/0475

Abstract: In various examples, reinforcement learning techniques are used during joint training of a generative model with at least one other model. For example, a first set of training data and a second set of training data generated by the generative model are combined and used to train an event detection model. In addition, in such examples, a reward is determined based on the performance of the event detection model (e.g., an agreement between gradients of a loss function of training data and synthetic data) and used at least in part to update the parameters of the generative model.