Abstract: The present disclosure describes a model for large scale color prediction of objects identified in images. Embodiments of the present disclosure include an object detection network, an attention network, and a color classification network. The object detection network generates object features for an object in an image and may include a convolutional neural network (CNN), region proposal network, or a ResNet. The attention network generates an attention vector for the object based on the object features, wherein the attention network takes a query vector based on the object features, and a plurality of key vector and a plurality of value vectors corresponding to a plurality of colors as input. The color classification network generates a color attribute vector based on the attention vector, wherein the color attribute vector indicates a probability of the object including each of the plurality of colors.

Abstract: Techniques and systems are provided for training a machine learning model using different datasets to perform one or more tasks. The machine learning model can include a first sub-module configured to perform a first task and a second sub-module configured to perform a second task. The first sub-module can be selected for training using a first training dataset based on a format of the first training dataset. The first sub-module can then be trained using the first training dataset to perform the first task. The second sub-module can be selected for training using a second training dataset based on a format of the second training dataset. The second sub-module can then be trained using the second training dataset to perform the second task.

Abstract: Systems and methods for color prediction are described. Embodiments of the present disclosure receive an image that includes an object including a color, generate a color vector based on the image using a color classification network, where the color vector includes a color value corresponding to each of a set of colors, generate a bias vector by comparing the color vector to teach of a set of center vectors, where each of the set of center vectors corresponds to a color of the set of colors, and generate an unbiased color vector based on the color vector and the bias vector, where the unbiased color vector indicates the color of the object.

Abstract: Systems and methods for intent discovery and video summarization are described. Embodiments of the present disclosure receive a video and a transcript of the video, encode the video to obtain a sequence of video encodings, encode the transcript to obtain a sequence of text encodings, apply a visual gate to the sequence of text encodings based on the sequence of video encodings to obtain gated text encodings, and generate an intent label for the transcript based on the gated text encodings.

Abstract: Multitask machine-learning model training and training data augmentation techniques are described. In one example, training is performed for multiple tasks simultaneously as part of training a multitask machine-learning model using question pairs. Examples of the multiple tasks include question summarization and recognizing question entailment. Further, a loss function is described that incorporates a parameter sharing loss that is configured to adjust an amount that parameters are shared between corresponding layers trained for the first and second tasks, respectively. In an implementation, training data augmentation techniques are also employed by synthesizing question pairs, automatically and without user intervention, to improve accuracy in model training.

Abstract: Systems and methods for text processing are described. Embodiments of the present disclosure receive a query comprising a natural language expression; extract a plurality of mentions from the query; generate a relation vector between a pair of the plurality of mentions using a relation encoder network, wherein the relation encoder network is trained using a contrastive learning process where mention pairs from a same document are labeled as positive samples and mention pairs from different documents are labeled as negative samples; combine the plurality of mentions with the relation vector to obtain a virtual knowledge graph of the query; identify a document corresponding to the query by comparing the virtual knowledge graph of the query to a virtual knowledge graph of the document; and transmit a response to the query, wherein the response includes a reference to the document.

Abstract: Systems and methods for coreference resolution are provided. One aspect of the systems and methods includes inserting a speaker tag into a transcript, wherein the speaker tag indicates that a name in the transcript corresponds to a speaker of a portion of the transcript; encoding a plurality of candidate spans from the transcript based at least in part on the speaker tag to obtain a plurality of span vectors; extracting a plurality of entity mentions from the transcript based on the plurality of span vectors, wherein each of the plurality of entity mentions corresponds to one of the plurality of candidate spans; and generating coreference information for the transcript based on the plurality of entity mentions, wherein the coreference information indicates that a pair of candidate spans of the plurality of candidate spans corresponds to a pair of entity mentions that refer to a same entity.

Abstract: Systems and methods for video segmentation and summarization are described. Embodiments of the present disclosure receive a video and a transcript of the video; generate visual features representing frames of the video using an image encoder; generate language features representing the transcript using a text encoder, wherein the image encoder and the text encoder are trained based on a correlation between training visual features and training language features; and segment the video into a plurality of video segments based on the visual features and the language features.

Abstract: The present invention is directed towards providing automated workflows for the identification of a reading order from text segments extracted from a document. Ordering the text segments is based on trained natural language models. In some embodiments, the workflows are enabled to perform a method for identifying a sequence associated with a portable document. The methods includes iteratively generating a probabilistic language model, receiving the portable document, and selectively extracting features (such as but not limited to text segments) from the document. The method may generate pairs of features (or feature pair from the extracted features). The method may further generate a score for each of the pairs based on the probabilistic language model and determine an order to features based on the scores. The method may provide the extracted features in the determined order.

Abstract: Systems and methods for natural language processing are described. Embodiments of the present disclosure identify a task set including a plurality of pseudo tasks, wherein each of the plurality of pseudo tasks includes a support set corresponding to a first natural language processing (NLP) task and a query set corresponding to a second NLP task; update a machine learning model in an inner loop based on the support set; update the machine learning model in an outer loop based on the query set; and perform the second NLP task using the machine learning model.

Abstract: Embodiments of the disclosure provide a machine learning model for generating a predicted executable command for an image. The learning model includes an interface configured to obtain an utterance indicating a request associated with the image, an utterance sub-model, a visual sub-model, an attention network, and a selection gate. The machine learning model generates a segment of the predicted executable command from weighted probabilities of each candidate token in a predetermined vocabulary determined based on the visual features, the concept features, current command features, and the utterance features extracted from the utterance or the image.

Abstract: Systems and methods for key-phrase extraction are described. The systems and methods include receiving a transcript including a text paragraph and generating key-phrase data for the text paragraph using a key-phrase extraction network. The key-phrase extraction network is trained to identify domain-relevant key-phrase data based on domain data obtained using a domain discriminator network. The systems and methods further include generating meta-data for the transcript based on the key-phrase data.

Abstract: Techniques for training a language model for code switching content are disclosed. Such techniques include, in some embodiments, generating a dataset, which includes identifying one or more portions within textual content in a first language, the identified one or more portions each including one or more of offensive content or non-offensive content; translating the identified one or more salient portions to a second language; and reintegrating the translated one or more portions into the textual content to generate code-switched textual content. In some cases, the textual content in the first language includes offensive content and non-offensive content, the identified one or more portions include the offensive content, and the translated one or more portions include a translated version of the offensive content. In some embodiments, the code-switched textual content is at least part of a synthetic dataset usable to train a language model, such as a multilingual classification model.

Abstract: Systems and methods for item recommendation are described. Embodiments of the present disclosure receive input indicating a relationship between a user and a first content item; generate a knowledge graph based on the input, wherein the knowledge graph comprises relationship information between the user and a plurality of content items; generate a first feature embedding representing the user and a second feature embedding representing a second content item of the plurality of content items based on the knowledge graph, wherein the second feature embedding is generated using a first modality for a query vector of an attention mechanism and a second modality for a key vector and a value vector of the attention mechanism; compare the first feature embedding to the second feature embedding to obtain a similarity score; and recommend the second content item for the user based on the similarity score.

Abstract: Systems and methods for image captioning are described. One or more aspects of the systems and methods include generating a training caption for a training image using an image captioning network; encoding the training caption using a multi-modal encoder to obtain an encoded training caption; encoding the training image using the multi-modal encoder to obtain an encoded training image; computing a reward function based on the encoded training caption and the encoded training image; and updating parameters of the image captioning network based on the reward function.

Abstract: Techniques for training a first neural network (NN) model using a pre-trained second NN model are disclosed. In an example, training data is input to the first and second models. The training data includes masked tokens and unmasked tokens. In response, the first model generates a first prediction associated with a masked token and a second prediction associated with an unmasked token, and the second model generates a third prediction associated with the masked token and a fourth prediction associated with the unmasked token. The first model is trained, based at least in part on the first, second, third, and fourth predictions. In another example, a prediction associated with a masked token, a prediction associated with an unmasked token, and a prediction associated with whether two sentences of training data are adjacent sentences are received from each of the first and second models. The first model is trained using the predictions.

Abstract: Systems and methods for natural language processing are described. One or more aspects of a method, apparatus, and non-transitory computer readable medium include receiving a text phrase; encoding the text phrase using an encoder to obtain a hidden representation of the text phrase, wherein the encoder is trained during a first training phrase using self-supervised learning based on a first contrastive loss and during a second training phrase using supervised learning based on a second contrastive learning loss; identifying an intent of the text phrase from a predetermined set of intent labels using a classification network, wherein the classification network is jointly trained with the encoder in the second training phase; and generating a response to the text phrase based on the intent.

Abstract: System and methods for a text summarization system are described. In one example, a text summarization system receives an input utterance and determines whether the utterance should be included in a summary of the text. The text summarization system includes an embedding network, a convolution network, an encoding component, and a summary component. The embedding network generates a semantic embedding of an utterance. The convolution network generates a plurality of feature vectors based on the semantic embedding. The encoding component identifies a plurality of latent codes respectively corresponding to the plurality of feature vectors. The summary component identifies a prominent code among the latent codes and to select the utterance as a summary utterance based on the prominent code.

Abstract: Methods and systems are provided for generating a customized speech recognition neural network system comprised of an adapted automatic speech recognition neural network and an adapted language model neural network. The automatic speech recognition neural network is first trained in a generic domain and then adapted to a target domain. The language model neural network is first trained in a generic domain and then adapted to a target domain. Such a customized speech recognition neural network system can be used to understand input vocal commands.

Abstract: The present disclosure describes a model for large scale color prediction of objects identified in images. Embodiments of the present disclosure include an object detection network, an attention network, and a color classification network. The object detection network generates object features for an object in an image and may include a convolutional neural network (CNN), region proposal network, or a ResNet. The attention network generates an attention vector for the object based on the object features, wherein the attention network takes a query vector based on the object features, and a plurality of key vector and a plurality of value vectors corresponding to a plurality of colors as input. The color classification network generates a color attribute vector based on the attention vector, wherein the color attribute vector indicates a probability of the object including each of the plurality of colors.