Abstract: A method for computer question answering includes, at a retriever subsystem of a question answering computer system, identifying a plurality of relevant text evidence strings for an input text question. At a linker subsystem of the question answering computer system, one or more of the plurality of relevant text evidence strings are associated with a respective secondary text evidence string to form a plurality of evidence chains via a previously-trained entity-linking machine-learning model. At a chainer subsystem of the question answering computer system, a ranked set of the evidence chains is identified based at least in part on an output of a generative machine-learning model applied to each of the plurality of evidence chains. At a reader subsystem of the question answering computer system, an answer to the input text question is output based at least in part on the ranked set of evidence chains.

Abstract: A computing system includes a logic subsystem and a storage subsystem holding instructions executable by the logic subsystem to implement a transformer-based text encoder. The transformer-based text encoder includes a plurality of transformer blocks previously-trained to apply encoding operations to computer-readable text representations of input text strings, the computer-readable text representations including computer-readable question representations of input text questions, and computer-readable passage representations of input text passages. The plurality of transformer blocks include a shared transformer block trained for both the computer-readable question representations and the computer-readable passage representations and a specialized transformer block including two or more input-specific subnetworks, and a routing function to select an input-specific subnetwork of the two or more input-specific subnetworks for each of the computer-readable text representations.

Abstract: Methods and systems are disclosed for improving dialog management for task-oriented dialog systems. The disclosed dialog builder leverages machine teaching processing to improve development of dialog managers. In this way, the dialog builder combines the strengths of both rule-based and machine-learned approaches to allow dialog authors to: (1) import a dialog graph developed using popular dialog composers, (2) convert the dialog graph to text-based training dialogs, (3) continuously improve the trained dialogs based on log dialogs, and (4) generate a corrected dialog for retraining the machine learning.

Abstract: Generally discussed herein are devices, systems, and methods for generating an embedding that is both local string dependent and global string dependent. The generated embedding can improve machine learning (ML) model performance. A method can include converting a string of words to a series of tokens, generating a local string-dependent embedding of each token of the series of tokens, generating a global string-dependent embedding of each token of the series of tokens, combining the local string dependent embedding the global string dependent embedding to generate an n-gram induced embedding of each token of the series of tokens, obtaining a masked language model (MLM) previously trained to generate a masked word prediction, and executing the MLM based on the n-gram induced embedding of each token to generate the masked word prediction.

Abstract: Systems and methods are provided for training and using a novel unified language foundation model. An encoder-decoder natural language model is obtained and various training data is obtained and used for training. The training process integrates a combination of replaced token detection, corrupted span reconstruction, and disentangled attention methodologies to produce a unified encoder-decoder model. The trained model is trained for performing both natural language understanding (NLU) tasks and natural language generation (NLG) tasks. Attention applied to the model is applied discretely to segmented chunks of encoded data during processing to improve the efficiency of applying attention by the model.

Abstract: Systems and methods are provided for training and using a novel unified language foundation model. An encoder-decoder natural language model is obtained and various training data is obtained and used for training. The training process integrates a combination of replaced token detection, corrupted span reconstruction, and disentangled attention methodologies to produce a unified encoder-decoder model. The trained model is trained for performing both natural language understanding (NLU) tasks and natural language generation (NLG) tasks. Attention applied to the model is applied discretely to segmented chunks of encoded data during processing to improve the efficiency of applying attention by the model.

Abstract: This document relates to machine learning. One example includes a method or technique that can be performed on a computing device. The method or technique can include obtaining a task-semantically-conditioned generative model that has been pretrained based at least on a first training data set having unlabeled training examples and semantically conditioned based at least on a second training data set having dialog act-labeled utterances. The method or technique can also include inputting dialog acts into the semantically-conditioned generative model and obtaining synthetic utterances that are output by the semantically-conditioned generative model. The method or technique can also include outputting the synthetic utterances.

Abstract: This document relates to training of machine learning models. One example method involves providing a machine learning model having one or more mapping layers. The one or more mapping layers can include at least a first mapping layer configured to map components of pretraining examples into first representations in a space. The example method also includes performing a pretraining stage on the one or more mapping layers using the pretraining examples. The pretraining stage can include adding noise to the first representations of the components of the pretraining examples to obtain noise-adjusted first representations. The pretraining stage can also include performing a self-supervised learning process to pretrain the one or more mapping layers using at least the first representations of the training data items and the noise-adjusted first representations of the training data items.

Abstract: A processing unit can train a model as a joint multi-domain recurrent neural network (JRNN), such as a bi-directional recurrent neural network (bRNN) and/or a recurrent neural network with long-short term memory (RNN-LSTM) for spoken language understanding (SLU). The processing unit can use the trained model to, e.g., jointly model slot filling, intent determination, and domain classification. The joint multi-domain model described herein can estimate a complete semantic frame per query, and the joint multi-domain model enables multi-task deep learning leveraging the data from multiple domains. The joint multi-domain recurrent neural (JRNN) can leverage semantic intents (such as, finding or identifying, e.g., a domain specific goal) and slots (such as, dates, times, locations, subjects, etc.) across multiple domains.

Abstract: Generally discussed herein are devices, systems, and methods for generating an embedding that is both local string dependent and global string dependent. The generated embedding can improve machine learning (ML) model performance. A method can include converting a string of words to a series of tokens, generating a local string-dependent embedding of each token of the series of tokens, generating a global string-dependent embedding of each token of the series of tokens, combining the local string dependent embedding the global string dependent embedding to generate an n-gram induced embedding of each token of the series of tokens, obtaining a masked language model (MLM) previously trained to generate a masked word prediction, and executing the MLM based on the n-based induced embedding of each token to generate the masked word prediction.

Abstract: This document relates to training of machine learning models. One example method involves providing a machine learning model having one or more mapping layers. The one or more mapping layers can include at least a first mapping layer configured to map components of pretraining examples into first representations in a space. The example method also includes performing a pretraining stage on the one or more mapping layers using the pretraining examples. The pretraining stage can include adding noise to the first representations of the components of the pretraining examples to obtain noise-adjusted first representations. The pretraining stage can also include performing a self-supervised learning process to pretrain the one or more mapping layers using at least the first representations of the training data items and the noise-adjusted first representations of the training data items.

Abstract: A processing unit can train a model as a joint multi-domain recurrent neural network (JRNN), such as a bi-directional recurrent neural network (bRNN) and/or a recurrent neural network with long-short term memory (RNN-LSTM) for spoken language understanding (SLU). The processing unit can use the trained model to, e.g., jointly model slot filling, intent determination, and domain classification. The joint multi-domain model described herein can estimate a complete semantic frame per query, and the joint multi-domain model enables multi-task deep learning leveraging the data from multiple domains. The joint multi-domain recurrent neural network (JRNN) can leverage semantic intents (such as, finding or identifying, e.g., a domain specific goal) and slots (such as, dates, times, locations, subjects, etc.) across multiple domains.

Abstract: Knowledge graphs can greatly improve the quality of content recommendation systems. There is a broad variety of knowledge graphs in the domain including clicked user-ad graphs, clicked query-ad graphs, keyword-display URL graphs etc. A hierarchical Transformer model learns entity embeddings in knowledge graphs. The model consists of two different Transformer blocks where the bottom block generates relation-dependent embeddings for the source entity and its neighbors, and the top block aggregates the outputs from the bottom block to produce the target entity embedding. To balance the information from contextual entities and the source entity itself, a masked entity model (MEM) task is combined with a link prediction task in model training.

Abstract: Generally discussed herein are devices, systems, and methods for scheduling tasks to be completed by resources. A method can include identifying features of the task, the features including a time-dependent feature and a time-independent feature, the time-dependent feature indicating a time the task is more likely to be successfully completed by the resource, converting the features to feature values based on a predefined mapping of features to feature values in a first memory device, determining, by a gradient boost tree model and based on a first current time and the feature values, a likelihood the resource will successfully complete the task, and scheduling the task to be performed by the resource based on the determined likelihood.

Abstract: Techniques for iterative query-based analysis of text are described. According to various implementations, a neural network architecture is implemented receives a query for information about text content, and iteratively analyzes the content using the query. During the analysis a state of the query evolves until it reaches a termination state, at which point the state of the query is output as an answer to the initial query.

Abstract: Systems and methods are provided for facilitating the building and use of natural language understanding models. The systems and methods identify a plurality of tokens and use them to generate one or more pre-trained natural language models using a transformer. The transformer disentangles the content embedding and positional embedding in the computation of its attention matrix. Systems and methods are also provided to facilitate self-training of the pre-trained natural language model by utilizing multi-step decoding to better reconstruct masked tokens and improve pre-training convergence.

Abstract: Knowledge graphs can greatly improve the quality of content recommendation systems. There is a broad variety of knowledge graphs in the domain including clicked user-ad graphs, clicked query-ad graphs, keyword-display URL graphs etc. A hierarchical Transformer model learns entity embeddings in knowledge graphs. The model consists of two different Transformer blocks where the bottom block generates relation-dependent embeddings for the source entity and its neighbors, and the top block aggregates the outputs from the bottom block to produce the target entity embedding. To balance the information from contextual entities and the source entity itself, a masked entity model (MEM) task is combined with a link prediction task in model training.

Abstract: A method for training a language model comprises (a) receiving vectorized training data as input to a multitask pretraining problem; (b) generating modified vectorized training data based on the vectorized training data, according to an upstream data embedding; (c) emitting pretraining output based on the modified vectorized training data, according to a downstream data embedding equivalent to the upstream data embedding; and (d) adjusting the upstream data embedding and the downstream data embedding by computing, based on the pretraining output, a gradient of the upstream data embedding disentangled from a gradient of the downstream data embedding, thereby advancing the multitask pretraining problem toward a pretrained state.

Abstract: Systems and methods are provided for determining a response to a query in a dialog. An entity extractor extracts rules and conditions associated with the query and determines a particular task. The disclosed technology generates a transformer-based dialog embedding by pre-training a transformer using dialog corpora including a plurality of tasks. A task-specific classifier generates a first set of candidate responses based on rules and conditions associated with the task. The transformer-based dialog embedding generates a second set of candidate responses to the query. The classifier accommodates changes made to a task by an interactive dialog editor as machine teaching. A response generator generates a response based on the first and second sets of candidate responses using an optimization function.

Abstract: A training technique trains a neural network having sparsely-activated sub-networks. It does so by processing plural batches of training data in two respective passes of the neural network, yielding first prediction information and second prediction information. For each batch, the technique randomly assigns different sub-networks in the first and second passes of the neural network to process the batch. Over the course of training, the technique attempts to minimize loss information, which describes the difference between the first prediction information and ground-truth information, and the difference between the second prediction information and the ground-truth information. Simultaneously, the technique attempts to minimize divergence information, which describes the divergence of the first prediction information from the second prediction information (and vice versa).