In recent years, the field of image generation has been revolutionized by the application of autoregressive transformers and DDPMs. These approaches model the process of image generation as a step-wise probabilistic processes and leverage large amounts of compute and data to learn the image distribution. This methodology of improving performance need not be confined to images. This paper describes a way to apply advances in the image generative domain to speech synthesis. The result is TorToise -- an expressive, multi-voice text-to-speech system. All model code and trained weights have been open-sourced at https://github.com/neonbjb/tortoise-tts.
In this work, we explore a scalable way for building a general representation model toward unlimited modalities. We release ONE-PEACE, a highly extensible model with 4B parameters that can seamlessly align and integrate representations across vision, audio, and language modalities. The architecture of ONE-PEACE comprises modality adapters, shared self-attention layers, and modality FFNs. This design allows for the easy extension of new modalities by adding adapters and FFNs, while also enabling multi-modal fusion through self-attention layers. To pretrain ONE-PEACE, we develop two modality-agnostic pretraining tasks, cross-modal aligning contrast and intra-modal denoising contrast, which align the semantic space of different modalities and capture fine-grained details within modalities concurrently. With the scaling-friendly architecture and pretraining tasks, ONE-PEACE has the potential to expand to unlimited modalities. Without using any vision or language pretrained model for initialization, ONE-PEACE achieves leading results on a wide range of uni-modal and multi-modal tasks, including image classification (ImageNet), semantic segmentation (ADE20K), audio-text retrieval (AudioCaps, Clotho), audio classification (ESC-50, FSD50K, VGGSound), audio question answering (AVQA), image-text retrieval (MSCOCO, Flickr30K), and visual grounding (RefCOCO/+/g). Code is available at https://github.com/OFA-Sys/ONE-PEACE.
During 2022, both transformer-based AI text generation sys-tems such as GPT-3 and AI text-to-image generation systems such as DALL-E 2 and Stable Diffusion made exponential leaps forward and are unquestionably altering the fields of digital art and electronic literature. In this panel a group of electronic literature authors and theorists consider new oppor-tunities for human creativity presented by these systems and present new works have produced during the past year that specifically address these systems as environments for literary expressions that are translated through iterative interlocutive processes into visual representations. The premise that binds these presentations is that these systems and the works gener-ated must be considered from a literary perspective, as they originate in human writing. In works ranging from a visual memoir of the personal experience of a health crisis, to interac-tive web comics, to architectures based on abstract poetic language, to political satire, four artists explore the capabili-ties of these writing environments for new genres of literary artist practice, while a digital culture theorist considers the origins and effects of the particular training datasets of human language and images on which these new hybrid forms are based.
We systematically investigate lightweight strategies to adapt large language models (LLMs) for the task of radiology report summarization (RRS). Specifically, we focus on domain adaptation via pretraining (on natural language, biomedical text, and clinical text) and via prompting (zero-shot, in-context learning) or parameter-efficient fine-tuning (prefix tuning, LoRA). Our results on the MIMIC-III dataset consistently demonstrate best performance by maximally adapting to the task via pretraining on clinical text and parameter-efficient fine-tuning on RRS examples. Importantly, this method fine-tunes a mere 0.32% of parameters throughout the model, in contrast to end-to-end fine-tuning (100% of parameters). Additionally, we study the effect of in-context examples and out-of-distribution (OOD) training before concluding with a radiologist reader study and qualitative analysis. Our findings highlight the importance of domain adaptation in RRS and provide valuable insights toward developing effective natural language processing solutions for clinical tasks.
In this paper, we introduce a new dataset of room interior pictures with overlaying and scene text, totalling to 4836 annotated images in 25 product categories. We provide details on the collection and annotation process of our dataset, and analyze its statistics. Furthermore, we propose a baseline method for overlaying text detection, that leverages the character region-aware text detection framework to guide the classification model. We validate our approach and show its efficiency in terms of binary classification metrics, reaching the final performance of 0.95 F1 score, with false positive and false negative rates of 0.02 and 0.06 correspondingly.
Cross-domain aspect-based sentiment analysis (ABSA) aims to perform various fine-grained sentiment analysis tasks on a target domain by transferring knowledge from a source domain. Since labeled data only exists in the source domain, a model is expected to bridge the domain gap for tackling cross-domain ABSA. Though domain adaptation methods have proven to be effective, most of them are based on a discriminative model, which needs to be specifically designed for different ABSA tasks. To offer a more general solution, we propose a unified bidirectional generative framework to tackle various cross-domain ABSA tasks. Specifically, our framework trains a generative model in both text-to-label and label-to-text directions. The former transforms each task into a unified format to learn domain-agnostic features, and the latter generates natural sentences from noisy labels for data augmentation, with which a more accurate model can be trained. To investigate the effectiveness and generality of our framework, we conduct extensive experiments on four cross-domain ABSA tasks and present new state-of-the-art results on all tasks. Our data and code are publicly available at \url{https://github.com/DAMO-NLP-SG/BGCA}.
We present SPECTRON, a novel approach to adapting pre-trained language models (LMs) to perform speech continuation. By leveraging pre-trained speech encoders, our model generates both text and speech outputs with the entire system being trained end-to-end operating directly on spectrograms. Training the entire model in the spectrogram domain simplifies our speech continuation system versus existing cascade methods which use discrete speech representations. We further show our method surpasses existing spoken language models both in semantic content and speaker preservation while also benefiting from the knowledge transferred from pre-existing models. Audio samples can be found in our website https://michelleramanovich.github.io/spectron/spectron
Over the years, there has been a paradigm shift in how users access financial services. With the advancement of digitalization more users have been preferring the online mode of performing financial activities. This has led to the generation of a huge volume of financial content. Most investors prefer to go through these contents before making decisions. Every industry has terms that are specific to the domain it operates in. Banking and Financial Services are not an exception to this. In order to fully comprehend these contents, one needs to have a thorough understanding of the financial terms. Getting a basic idea about a term becomes easy when it is explained with the help of the broad category to which it belongs. This broad category is referred to as hypernym. For example, "bond" is a hypernym of the financial term "alternative debenture". In this paper, we propose a system capable of extracting and ranking hypernyms for a given financial term. The system has been trained with financial text corpora obtained from various sources like DBpedia [4], Investopedia, Financial Industry Business Ontology (FIBO), prospectus and so on. Embeddings of these terms have been extracted using FinBERT [3], FinISH [1] and fine-tuned using SentenceBERT [54]. A novel approach has been used to augment the training set with negative samples. It uses the hierarchy present in FIBO. Finally, we benchmark the system performance with that of the existing ones. We establish that it performs better than the existing ones and is also scalable.
We investigate different natural language processing (NLP) approaches based on contextualised word representations for the problem of early prediction of lung cancer using free-text patient medical notes of Dutch primary care physicians. Because lung cancer has a low prevalence in primary care, we also address the problem of classification under highly imbalanced classes. Specifically, we use large Transformer-based pretrained language models (PLMs) and investigate: 1) how \textit{soft prompt-tuning} -- an NLP technique used to adapt PLMs using small amounts of training data -- compares to standard model fine-tuning; 2) whether simpler static word embedding models (WEMs) can be more robust compared to PLMs in highly imbalanced settings; and 3) how models fare when trained on notes from a small number of patients. We find that 1) soft-prompt tuning is an efficient alternative to standard model fine-tuning; 2) PLMs show better discrimination but worse calibration compared to simpler static word embedding models as the classification problem becomes more imbalanced; and 3) results when training models on small number of patients are mixed and show no clear differences between PLMs and WEMs. All our code is available open source in \url{https://bitbucket.org/aumc-kik/prompt_tuning_cancer_prediction/}.
Subword segmenters like BPE operate as a preprocessing step in neural machine translation and other (conditional) language models. They are applied to datasets before training, so translation or text generation quality relies on the quality of segmentations. We propose a departure from this paradigm, called subword segmental machine translation (SSMT). SSMT unifies subword segmentation and MT in a single trainable model. It learns to segment target sentence words while jointly learning to generate target sentences. To use SSMT during inference we propose dynamic decoding, a text generation algorithm that adapts segmentations as it generates translations. Experiments across 6 translation directions show that SSMT improves chrF scores for morphologically rich agglutinative languages. Gains are strongest in the very low-resource scenario. SSMT also learns subwords that are closer to morphemes compared to baselines and proves more robust on a test set constructed for evaluating morphological compositional generalisation.