Abstract: Embodiments are disclosed for user-guided variable-rate compression. A method of user-guided variable-rate compression includes receiving a request to compress an image, the request including the image, a corresponding importance data, and a target bitrate, providing the image, the corresponding importance data, and the target bitrate to a compression network, generating, by the compression network, a learned importance map and a representation of the image, and generating, by the compressing network, a compressed representation of the image based on the learned importance map and the representation of the image.

Abstract: Embodiments of the present invention provide systems, methods, and computer storage media for pre-training entity extraction models to facilitate domain adaptation in resource-constrained domains. In an example embodiment, a first machine learning model is used to encode sentences of a source domain corpus and a target domain corpus into sentence embeddings. The sentence embeddings of the target domain corpus are combined into a target corpus embedding. Training sentences from the source domain corpus within a threshold of similarity to the target corpus embedding are selected. A second machine learning model is trained on the training sentences selected from the source domain corpus.

Abstract: A text style transfer system is described that generates different stylized versions of input text by rewriting the input text according to a target style. To do so, the text style transfer system employs a variational autoencoder to derive separate content and style representations for the input text, where the content representation specifies semantic information conveyed by the input text and the style representation specifies one or more style attributes expressed by the input text. The style representation using counterfactual reasoning to identify different transfer strengths for applying the target style to the input text. Each transfer strength represents a minimum change to the input text that achieves a different expression of the target style. The transfer strengths are then used to generate style representation variants, which are each concatenated with the content representation of the input text to generate the plurality of different stylized versions of the input text.

Abstract: A text summarization system auto-generates text summarization models using a combination of neural architecture search and knowledge distillation. Given an input dataset for generating/training a text summarization model, neural architecture search is used to sample a search space to select a network architecture for the text summarization model. Knowledge distillation includes fine-tuning a language model for a given text summarization task using the input dataset, and using the fine-tuned language model as a teacher model to inform the selection of the network architecture and the training of the text summarization model. Once a text summarization model has been generated, the text summarization model can be used to generate summaries for given text.