公开(公告)号：WO2022225806A1

公开(公告)日：2022-10-27

申请号：PCT/US2022/024995

申请日：2022-04-15

Applicant: NEC LABORATORIES AMERICA, INC.

Inventor： LIU, Yanchi ,  CHEN, Haifeng ,  ZHANG, Xuchao

IPC: G06F16/35 ,  G06F40/279 ,  G06K9/62 ,  G06N3/08

Abstract: Systems and methods are provided for detail matching. The method includes training a feature classifier (200) to identify technical features, and training a neural network model for a trained importance calculator (300) to calculate an importance value for each identified technical feature. The method further includes receiving a specification sheet (110) including a plurality of technical features, and receiving a plurality of descriptive sheets (120) each including a plurality of technical features. The method further includes identifying the technical features (130) in the specification sheet and the plurality of descriptive sheets using the trained feature classifier (200), and calculating an importance (140) for each identified technical feature using the trained feature importance calculator (300). The method further includes calculating a matching score (150) between the identified technical features of the specification sheet and the identified technical features of the plurality of descriptive sheets based on the importance of each identified technical feature.

公开(公告)号：WO2022169656A1

公开(公告)日：2022-08-11

申请号：PCT/US2022/013982

申请日：2022-01-27

Applicant: NEC LABORATORIES AMERICA, INC.

Inventor： LIU, Yanchi ,  ZONG, Bo ,  CHEN, Haifeng ,  ZHANG, Xuchao ,  ZHANG, Denghui

IPC: G06N5/02 ,  G06N20/00 ,  G06F40/211

Abstract: A method for employing a knowledge-driven pre-training framework for learning product representation is presented. The method includes learning (1001) contextual semantics of a product domain by a language acquisition stage including a context encoder and two language acquisition tasks, obtaining (1003) multi-faceted product knowledge by a knowledge acquisition stage including a knowledge encoder, skeleton attention layers, and three heterogeneous embedding guided knowledge acquisition tasks, generating (1005) local product representations defined as knowledge copies (KC) each capturing one facet of the multi-faceted product knowledge, and generating (1007) final product representation during a fine-tuning stage by combining all the KCs through a gating network.

公开(公告)号：WO2022245590A1

公开(公告)日：2022-11-24

申请号：PCT/US2022/028535

申请日：2022-05-10

Applicant: NEC LABORATORIES AMERICA, INC.

Inventor： ZHANG, Xuchao ,  CHEN, Haifeng ,  CHENG, Wei

IPC: G06F8/72 ,  G06F8/41 ,  G06N3/04 ,  G06N3/08 ,  G06F40/58

Abstract: Systems and methods are provided for automated computer code editing. The method includes training a code-editing neural network model (320) using a corpus of code editing data samples, including the pre-editing samples (110) and post-editing samples (120), and parsing (130) the pre-editing samples and post-editing samples into an Abstract Syntax Tree (AST). The method further includes using a grammar specification to transform (130) the AST tree into a unified Abstract Syntax Description Language (ASDL) graph for different programming languages, and using a gated graph neural network (GGNN) (320) to compute a vector representation (140, 150) for each node in the unified Abstract Syntax Description Language (ASDL) graph. The method further includes selecting and aggregating (160) support samples based on a query code with a multi-extent ensemble method, and altering the query code (170) iteratively using the pattern learned from the pre- and post-editing samples.

公开(公告)号：WO2022240558A1

公开(公告)日：2022-11-17

申请号：PCT/US2022/025487

申请日：2022-04-20

Applicant: NEC LABORATORIES AMERICA, INC.

Inventor： CHENG, Wei ,  CHEN, Haifeng ,  ZHANG, Xuchao ,  LUO, Dongsheng

IPC: G06F16/34 ,  G06F40/279 ,  G06F40/169 ,  G06N3/08 ,  G06N3/04 ,  G06F16/345 ,  G06F40/284 ,  G06F40/289

Abstract: A computer-implemented method is provided for keyphrase generation. The method includes pretraining (1210), by a processor device, a policy neural network on training documents using a sequence-to-sequence model. The training documents are each associated with a list of keyphrases included therein. The method further includes training (1220), by the processor device, the policy neural network using reinforcement learning with a summarization reward on present annotated keyphrases in an input training document and absent annotated keyphrase from the input training document that semantically describe a concept of the input training document. The method also includes predicting (1230), by the processor device, new keyphrases using the trained policy neural network.

公开(公告)号：WO2023059580A1

公开(公告)日：2023-04-13

申请号：PCT/US2022/045602

申请日：2022-10-04

Applicant: NEC LABORATORIES AMERICA, INC.

Inventor： CHEN, Haifeng ,  YU, Wenchao ,  CHEN, Yuncong ,  CHEN, Zhengzhang ,  ZHANG, Xuchao ,  TANG, LuAn ,  HE, Zexue

IPC: G06F40/58 ,  G06F40/56 ,  G10L15/26 ,  G10L15/00 ,  G06F3/16 ,  G06N20/00 ,  G06N3/04 ,  G06F40/30 ,  G06F40/47 ,  G06N3/08

Abstract: A computer-implemented method for multi-model representation learning is provided. The method includes encoding, by a trained time series (TS) encoder, an input TS segment into a TS-shared latent representation and a TS-private latent representation. The method further includes generating, by a trained text generator, a natural language text that explains the input TS segment, responsive to the TS-shared latent representation, the TS-private latent representation, and a text-private latent representation.

公开(公告)号：WO2022240557A1

公开(公告)日：2022-11-17

申请号：PCT/US2022/025481

申请日：2022-04-20

Applicant: NEC LABORATORIES AMERICA, INC.

Inventor： ZHANG, Xuchao ,  CHEN, Haifeng

IPC: G06F40/58 ,  G06F40/47 ,  G06N3/08 ,  G06N20/00 ,  G06F40/20 ,  G06F40/211 ,  G06F40/295 ,  G06F40/56 ,  G06N5/046

Abstract: A method provided for cross-lingual transfer trains (230) a pre-trained multi-lingual language model based on a gold labeled training set in a source language to obtain a trained model. The method assigns (240) each sample in an unlabeled target language set to a silver label according to a model prediction by the trained model to obtain set of silver labels, and performs (250) uncertainty-aware label selection based on the silver label assigned to each sample according to the model prediction and the trained model to obtain selected silver labels. The method performs iterative training (260) on the selected labels by applying the selected silver labels in the target language set as training labels and re-training the trained model with the gold labels and the selected silver labels to obtain an iterative model, and performs (270) task-specific result prediction in target languages based on the iterative model to generate a final predicted result in target languages.