Salesforce Com, Inc.

Salesforce, Inc.

Nitish Shirish Keskar

Caiming Xiong

Bryan McCann

Richard Socher

Chu Hong Hoi

Wenhao Liu

Ali Madani

Nikhil Naik

Jasdeep Singh

Kevin Zhang

Yingbo Zhou

Tian Xie

Jean-Marc Soumet

Roojuta Lalani

Thomas Archie Cook, Jr

Xinyi Yang

Huan Wang

Chen Xing

Semih Yavuz

Manju Vijayakumar

Tong Niu

Michael Machado

Johnson Liu

Juan Rodriguez

Qiong Liu

Rafael Amsili

Karim Ahmed

Karl Ryszard Skucha

James Douglas Harrison

Amit Hetawal

Vishal Motwani

Zhihao Zou

Minal Khodani

Adarsha Badarinath

Kevin Schoen

Chitra Gulabrani

Srikanth Kollu

Rohiniben Thakar

Bryan Mccann

Embodiments are directed to pre-training a transformer model using more parameters for sophisticated patterns (PSP++). The transformer model is divided into a held-out model and a main model. A forward pass and a backward pass are performed on the held-out model, where the forward pass determines self-attention hidden states of the held-out model and the backward pass determines loss of the held-out model. A forward pass on the main model is performed to determine a self-attention hidden states of the main model. The self-attention hidden states of the main model are concatenated with the self-attention hidden states of the held-out model. A backward pass is performed on the main model to determine a loss of the main model. The parameters of the held-out model are updated to reflect the loss of the held-out model and parameters of the main model are updated to reflect the loss of the main model.

Parameter utilization for language pre-training

A method for sequence-to-sequence prediction using a neural network model includes generating an encoded representation based on an input sequence using an encoder of the neural network model and predicting an output sequence based on the encoded representation using a decoder of the neural network model. The neural network model includes a plurality of model parameters learned according to a machine learning process. At least one of the encoder or the decoder includes a branched attention layer. Each branch of the branched attention layer includes an interdependent scaling node configured to scale an intermediate representation of the branch by a learned scaling parameter. The learned scaling parameter depends on one or more other learned scaling parameters of one or more other interdependent scaling nodes of one or more other branches of the branched attention layer.

Sequence-to-sequence prediction using a neural network model

Approaches for cross-lingual regularization for multilingual generalization include a method for training a natural language processing (NLP) deep learning module. The method includes accessing a first dataset having a first training data entry, the first training data entry including one or more natural language input text strings in a first language; translating at least one of the one or more natural language input text strings of the first training data entry from the first language to a second language; creating a second training data entry by starting with the first training data entry and substituting the at least one of the natural language input text strings in the first language with the translation of the at least one of the natural language input text strings in the second language; adding the second training data entry to a second dataset; and training the deep learning module using the second dataset.

Cross-lingual regularization for multilingual generalization

Embodiments described herein provide dynamic blocking, a decoding algorithm which enables large-scale pretrained language models to generate high-quality paraphrases in an un-supervised setting. Specifically, in order to obtain an alternative surface form, when the language model emits a token that is present in the source sequence, the language model is prevented from generating the next token that is the same as the subsequent source token in the source sequence at the next time step. In this way, the language model is forced to generate a paraphrased sequence of the input source sequence, but with mostly different wording.

Systems and methods for unsupervised paraphrase generation

An analytics system uses one or more machine-learned models to predict a hormone receptor status from a H&E stain image. The system partitions H&E stain images each into a plurality of non-overlapping image tiles. Bags of tiles are created through sampling of the image tiles. For each H&E stain image, the system generates a feature vector from a bag of tiles sampled from the partitioned image tiles. The analytics system trains one or more machine-learned models with training H&E stain images having a positive or negative receptor status. With the trained models, the analytics system predicts a hormone receptor status by applying a prediction model to the feature vector for a test H&E stain image.

Machine-learned hormone status prediction from image analysis

A multi-tenant system performs custom configuration of a tenant-specific chatbot to process and act upon natural language requests. The multi-tenant system configures the tenant-specific chatbots without requiring tenant-specific training. The multi-tenant system providing a user interface for configuring a tenant-specific set of permitted actions. The multi-tenant system determines a set of example phrases for each of the selected permitted actions. The multi-tenant system receives a natural language request from a user and identifies the action that the user wants to perform. The multi-tenant system uses a neural network to compare the natural language request with example phrases to identify an example phrase that matches the natural language request. The multi-tenant system performs the action corresponding to the matching example phrase.

Machine learning based tenant-specific chatbots for performing actions in a multi-tenant system

Systems and methods for unifying question answering and text classification via span extraction include a preprocessor for preparing a source text and an auxiliary text based on a task type of a natural language processing task, an encoder for receiving the source text and the auxiliary text from the preprocessor and generating an encoded representation of a combination of the source text and the auxiliary text, and a span-extractive decoder for receiving the encoded representation and identifying a span of text within the source text that is a result of the NLP task. The task type is one of entailment, classification, or regression. In some embodiments, the source text includes one or more of text received as input when the task type is entailment, a list of classifications when the task type is entailment or classification, or a list of similarity options when the task type is regression.

Systems and methods for unifying question answering and text classification via span extraction

Approaches for multitask learning as question answering include an input layer for encoding a context and a question, a self-attention based transformer including an encoder and a decoder, a first bi-directional long-term short-term memory (biLSTM) for further encoding an output of the encoder, a long-term short-term memory (LSTM) for generating a context-adjusted hidden state from the output of the decoder and a hidden state, an attention network for generating first attention weights based on an output of the first biLSTM and an output of the LSTM, a vocabulary layer for generating a distribution over a vocabulary, a context layer for generating a distribution over the context, and a switch for generating a weighting between the distributions over the vocabulary and the context, generating a composite distribution based on the weighting, and selecting a word of an answer using the composite distribution.

Multitask learning as question answering

Approaches for natural language processing include a multi-layer encoder for encoding words from a context and words from a question in parallel, a multi-layer decoder for decoding the encoded context and the encoded question, a pointer generator for generating distributions over the words from the context, the words from the question, and words in a vocabulary based on an output from the decoder, and a switch. The switch generates a weighting of the distributions over the words from the context, the words from the question, and the words in the vocabulary, generates a composite distribution based on the weighting of the distribution over the first words from the context, the distribution over the second words from the question, and the distribution over the words in the vocabulary, and selects words for inclusion in an answer using the composite distribution.

An analytics system uses one or more machine-learned models to predict a hormone receptor status from a H&E stain image. The system partitions H&E stain images each into a plurality of image tiles. Bags of tiles are created through sampling of the image tiles. The analytics system trains one or more machine-learned models with training H&E stain images having a positive or negative receptor status. The analytics system generates, via a tile featurization model, a tile feature vector for each image tile a test bag for a test H&E stain image. The analytics system generates, via an attention model, an aggregate feature vector for the test bag by aggregating the tile feature vectors of the test bag, wherein an attention weight is determined for each tile feature vector. The analytics system predicts a hormone receptor status by applying a prediction model to the aggregate feature vector for the test bag.

Approaches for multitask learning as question answering include a method for training that includes receiving a plurality of training samples including training samples from a plurality of task types, presenting the training samples to a neural model to generate an answer, determining an error between the generated answer and the natural language ground truth answer for each training sample presented, and adjusting parameters of the neural model based on the error. Each of the training samples includes a natural language context, question, and ground truth answer. An order in which the training samples are presented to the neural model includes initially selecting the training samples according to a first training strategy and switching to selecting the training samples according to a second training strategy. In some embodiments the first training strategy is a sequential training strategy and the second training strategy is a joint training strategy.

Hybrid training of deep networks includes a multi-layer neural network. The training includes setting a current learning algorithm for the multi-layer neural network to a first learning algorithm. The training further includes iteratively applying training data to the neural network, determining a gradient for parameters of the neural network based on the applying of the training data, updating the parameters based on the current learning algorithm, and determining whether the current learning algorithm should be switched to a second learning algorithm based on the updating. The training further includes, in response to the determining that the current learning algorithm should be switched to a second learning algorithm, changing the current learning algorithm to the second learning algorithm and initializing a learning rate of the second learning algorithm based on the gradient and a step used by the first learning algorithm to update the parameters of the neural network.

Hybrid training of deep networks

Growing community of inventors

San Bruno, CA, United States of America

Nitish Shirish Keskar