Abstract: Techniques for augmentation and batch balancing of training data to enhance negation and fairness of a machine learning model. In one particular aspect, a method is provided that includes obtaining a training set of labeled examples for training a machine learning model to classify sentiment, searching the training set of labeled examples or an unlabeled corpus of text on target domains for sentiment examples having negation cues, sentiment laden words, words with sentiment prefixes or suffixes, or a combination thereof, rewriting the sentiment examples to create negated versions thereof and generate a labeled negation pair data set, and training the machine learning model using labeled examples from the labeled negation pair data set.

Abstract: Techniques for augmentation and batch balancing of training data to enhance negation and fairness of a machine learning model. In one particular aspect, a method is provided that includes generating a list of demographic words associated with a demographic group, searching an unlabeled corpus of text to identify unlabeled examples in a target domain comprising at least one demographic word from the list of demographic words, rewriting the unlabeled examples to create one or more versions of each of the unlabeled examples and generate a fairness invariance data set, and training the machine learning model using unlabeled examples from the fairness invariance data set.

Abstract: Techniques for stop word data augmentation for training chatbot systems in natural language processing. In one particular aspect, a computer-implemented method includes receiving a training set of utterances for training an intent classifier to identify one or more intents for one or more utterances; augmenting the training set of utterances with stop words to generate an augmented training set of out-of-domain utterances for an unresolved intent category corresponding to an unresolved intent; and training the intent classifier using the training set of utterances and the augmented training set of out-of-domain utterances. The augmenting includes: selecting one or more utterances from the training set of utterances, and for each selected utterance, preserving existing stop words within the utterance and replacing at least one non-stop word within the utterance with a stop word or stop word phrase selected from a list of stop words to generate an out-of-domain utterance.

Abstract: Techniques disclosed herein relate generally to language detection. In one particular aspect, a method is provided that includes obtaining a sequence of n-grams of a textual unit; using an embedding layer to obtain an ordered plurality of embedding vectors for the sequence of n-grams; using a deep network to obtain an encoded vector that is based on the ordered plurality of embedding vectors; and using a classifier to obtain a language prediction for the textual unit that is based on the encoded vector. The deep network includes an attention mechanism, and using the embedding layer to obtain the ordered plurality of embedding vectors comprises, for each n-gram in the sequence of n-grams: obtaining hash values for the n-gram; based on the hash values, selecting component vectors from among the plurality of component vectors; and obtaining an embedding vector for the n-gram that is based on the component vectors.

Abstract: In some aspects, a computer obtains a trained conditional random field (CRF) model comprising a set of model parameters learned from training data and stored in a transition matrix. Tag sequences, inconsistent with the tag sequence logic, are identified for the tags within the transition matrix. setting, within the transition matrix, a cost associated with transitioning between the pair of tags to be equal to a predefined hyperparameter value that penalizes the transitioning between the inconsistent pair of tags. The CRF model receives a string of text comprising one or more named entities. The CRF model inputs the string of text into the CRF model having the cost associated with the transitioning between the pair of tags set equal to the predefined hyperparameter value. The CRF model classifies the words within the string of text into different classes which might include the one or more named entities.

Abstract: Deep learning techniques are disclosed for extraction of embedded data from documents. In an exemplary technique, a set of unstructured text data is received. One or more text groupings are generated by processing the set of unstructured text data. One or more text grouping embeddings are generated in a format for input to a machine learning model based on the one or more generated text groupings. One or more output predictions are generated by inputting the one or more text grouping embeddings into the machine learning model. Each output prediction of the one or more output predictions correspond to a predicted aspect of a text grouping of the one or more text groupings.

Abstract: Techniques are provided for customizing or fine-tuning a pre-trained version of a machine-learning model that includes multiple layers and is configured to process audio or textual language input. Each of the multiple layers is configured with a plurality of layer-specific pre-trained parameter values corresponding to a plurality of parameters, and each of the multiple layers is configured to implement multi-head attention. An incomplete subset of the multiple layers is identified for which corresponding layer-specific pre-trained parameter values are to be fine-tuned using a client data set. The machine-learning model is fine-tuned using the client data set to generate an updated version of the machine-learning model, where the layer-specific pre-trained parameter values configured for each layer of one of more of the multiple layers not included in the incomplete subset are frozen during the fine-tuning. Use of the updated version of the machine-learning model is facilitated.

Abstract: Techniques for extracting key information from a document using machine-learning models in a chatbot system is disclosed herein. In one particular aspect, a method is provided that includes receiving a set of data, which includes key fields, within a document at a data processing system that includes a table detection module, a key information extraction module, and a table extraction module. Text information and corresponding location data are extracted via optical character recognition. The table detection module detects whether one or more tables are present in the document and, if applicable, a location of each of the tables. The key information extraction module extracts text from the key fields. The table extraction module extracts each of the tables based on input from the optical character recognition and the table detection module. Extraction results include the text from the key fields and each of the tables can be output.

Abstract: Techniques are disclosed herein for focused training of language models and end-to-end hypertuning of the framework. In one aspect, a method is provided that includes obtaining a machine learning model pre-trained for language modeling, and post-training the machine learning model for various tasks to generate a focused machine learning model. The post-training includes: (i) training the machine learning model on an unlabeled set of training data pertaining to a task that the machine learning model was pre-trained for as part of the language modeling, and the unlabeled set of training data is obtained with respect to a target domain, a target task, or a target language, and (ii) training the machine learning model on a labeled set of training data that pertains to another task that is an auxiliary task related to a downstream task to be performed using the machine learning model or output from the machine learning model.

Abstract: Techniques disclosed herein relate generally to text classification and include techniques for fusing word embeddings with word scores for text classification. In one particular aspect, a method for text classification is provided that includes obtaining an embedding vector for a textual unit, based on a plurality of word embedding vectors and a plurality of word scores. The plurality of word embedding vectors includes a corresponding word embedding vector for each of a plurality of words of the textual unit, and the plurality of word scores includes a corresponding word score for each of the plurality of words of the textual unit. The method also includes passing the embedding vector for the textual unit through at least one feed-forward layer to obtain a final layer output, and performing a classification on the final layer output.

Abstract: Techniques for adjusting outlier datasets for training chatbot systems in natural language processing are disclosed. In one particular aspect, a method is provided that includes receiving a dataset that includes training or inference data. An initial set of outlier data points can be identified within the dataset based on a score of the outlier data points being above or below a threshold. The initial set can be adjusted by identifying one or more nearest neighbors, which can be included in the dataset. Outlier data points that include a label that matches a number of labels of the nearest neighbors that exceeds a predetermined threshold can be removed from the initial set of outlier data points to generate a final set. Outlier data points of the final set can be adjusted with respect to the dataset to generate a set of training data that is used to train a machine-learning model.

Abstract: Techniques for improving a semantic parser of a dialog system, by breaking the semantic parser into a coarse semantic parser and a fine semantic parser, are described. A method described herein includes accessing an utterance received in a dialog system. The utterance is a text-based natural language expression. The method further includes applying a coarse semantic parser to the utterance to determine an intermediate logical form for the utterance. The intermediate logical form indicates one or more intents in the utterance. The method further includes applying a fine semantic parser to the intermediate logical form to determine a logical form for the utterance. The logical form is a syntactic expression of the utterance according to an established grammar, and the logical form includes one or more parameters of the one or more intents. The logical form can be used to conduct a dialog with a user of the dialog system.

Abstract: Techniques described herein use backpropagation to train one or more machine learning (ML) models of a dialog system. For instance, a method includes accessing seed data that includes training tuples, where each training tuple comprising a respective logical form. The method includes converting the logical form of a training tuple to a converted logical form, by applying to the logical form a text-to-speech (TTS) subsystem, an automatic speech recognition (ASR) subsystem, and a semantic parser of a dialog system. The method includes determining a training signal by using an objective function to compare the converted logical form to the logical form. The method further includes training the TTS subsystem, the ASR subsystem, and the semantic parser via backpropagation based on the training signal. As a result of the training by backpropagation, the machine learning models are tuned work effectively together within a pipeline of the dialog system.

Abstract: Described herein are dialog systems, and techniques for providing such dialog systems, that are suitable for use on standalone computing devices. In some embodiments, a dialog system includes a dialog manager, which takes as input an input logical form, which may be a representation of user input. The dialog, manager may include a dialog state tracker, an execution subsystem, a dialog policy subsystem, and a context stack. The dialog state tracker may generate an intermediate logical form from the input logical form combined with a context from the context stack. The context stack may maintain a history of a current dialog, and thus, the intermediate logical form may include contextual information potentially missing from the input logical form. The execution subsystem may execute the intermediate logical form to produce an execution result, and the dialog policy subsystem may generate an output logical form based on the execution result.

Abstract: Techniques are described for using data stored for a user in association with context levels to improve the efficiency and accuracy of dialog processing tasks. A dialog system stores historical dialog data in association with a plurality of configured context levels. The dialog system receives an utterance and identifies a term for disambiguation from the utterance. Based on a determined context level, the dialog system identifies relevant historical data stored to a database. The historical data may be used to perform tasks such as resolving an ambiguity based on user preferences, disambiguating named entities based on a prior dialog, and identifying previously generated answers to queries. Based on the context level, the dialog system can efficiently identify the relevant information and use the identified information to provide a response.

Abstract: Techniques for noise data augmentation for training chatbot systems in natural language processing. In one particular aspect, a method is provided that includes receiving a training set of utterances for training an intent classifier to identify one or more intents for one or more utterances; augmenting the training set of utterances with noise text to generate an augmented training set of utterances; and training the intent classifier using the augmented training set of utterances. The augmenting includes: obtaining the noise text from a list of words, a text corpus, a publication, a dictionary, or any combination thereof irrelevant of original text within the utterances of the training set of utterances, and incorporating the noise text within the utterances relative to the original text in the utterances of the training set of utterances at a predefined augmentation ratio to generate augmented utterances.

Abstract: Techniques are described for training and executing a machine learning model using data derived from a database. A dialog system uses data from the database to generate related training data for natural language understanding applications. The generated training data is then used to train a machine learning model. This enables the dialog system to leverage a large amount of available data to speed up the training process as compared to conventional labeling techniques. The dialog system uses the trained machine learning model to identify a named entity from a received spoken utterance and generate and output a speech response based upon the identified named entity.

Abstract: Techniques described herein use backpropagation to train one or more machine learning (ML) models of a dialog system. For instance, a method includes accessing seed data that includes training tuples, where each training tuple comprising a respective logical form. The method includes converting the logical form of a training tuple to a converted logical form, by applying to the logical form a text-to-speech (TTS) subsystem, an automatic speech recognition (ASR) subsystem, and a semantic parser of a dialog system. The method includes determining a training signal by using an objective function to compare the converted logical form to the logical form. The method further includes training the TTS subsystem, the ASR subsystem, and the semantic parser via backpropagation based on the training signal. As a result of the training by backpropagation, the machine learning models are tuned work effectively together within a pipeline of the dialog system.

Abstract: Techniques for improving a semantic parser of a dialog system, by breaking the semantic parser into a coarse semantic parser and a fine semantic parser, are described. A method described herein includes accessing an utterance received in a dialog system. The utterance is a text-based natural language expression. The method further includes applying a coarse semantic parser to the utterance to determine an intermediate logical form for the utterance. The intermediate logical form indicates one or more intents in the utterance. The method further includes applying a fine semantic parser to the intermediate logical form to determine a logical form for the utterance. The logical form is a syntactic expression of the utterance according to an established grammar, and the logical form includes one or more parameters of the one or more intents. The logical form can be used to conduct a dialog with a user of the dialog system.

Abstract: Some techniques described herein determine a correction model for a dialog system, such that the correction model corrects output from an automatic speech recognition (ASR) subsystem in the dialog system. A method described herein includes accessing training data. A first tuple of the training data includes an utterance, where the utterance is a textual representation of speech. The method further includes using an ASR subsystem of a dialog system to convert the utterance to an output utterance. The method further includes storing the output utterance in corrective training data that is based on the training data. The method further includes training a correction model based on the corrective training data, such that the correction model is configured to correct output from the ASR subsystem during operation of the dialog system.