In healthcare, multimodal data is prevalent and requires to be comprehensively analyzed before diagnostic decisions, including medical images, clinical reports, etc. However, current large-scale artificial intelligence models predominantly focus on single-modal cognitive abilities and neglect the integration of multiple modalities. Therefore, we propose Stone Needle, a general multimodal large-scale model framework tailored explicitly for healthcare applications. Stone Needle serves as a comprehensive medical multimodal model foundation, integrating various modalities such as text, images, videos, and audio to surpass the limitations of single-modal systems. Through the framework components of intent analysis, medical foundation models, prompt manager, and medical language module, our architecture can perform multi-modal interaction in multiple rounds of dialogue. Our method is a general multimodal large-scale model framework, integrating diverse modalities and allowing us to tailor for specific tasks. The experimental results demonstrate the superior performance of our method compared to single-modal systems. The fusion of different modalities and the ability to process complex medical information in Stone Needle benefits accurate diagnosis, treatment recommendations, and patient care.
The prodigious growth of digital health data has precipitated a mounting interest in harnessing machine learning methodologies, such as natural language processing (NLP), to scrutinize medical records, clinical notes, and other text-based health information. Although NLP techniques have exhibited substantial potential in augmenting patient care and informing clinical decision-making, data privacy and adherence to regulations persist as critical concerns. Federated learning (FL) emerges as a viable solution, empowering multiple organizations to train machine learning models collaboratively without disseminating raw data. This paper proffers a pragmatic approach to medical NLP by amalgamating FL, NLP models, and the NVFlare framework, developed by NVIDIA. We introduce two exemplary NLP models, the Long-Short Term Memory (LSTM)-based model and Bidirectional Encoder Representations from Transformers (BERT), which have demonstrated exceptional performance in comprehending context and semantics within medical data. This paper encompasses the development of an integrated framework that addresses data privacy and regulatory compliance challenges while maintaining elevated accuracy and performance, incorporating BERT pretraining, and comprehensively substantiating the efficacy of the proposed approach.
Recently, there has been increasing progress in end-to-end automatic speech recognition (ASR) architecture, which transcribes speech to text without any pre-trained alignments. One popular end-to-end approach is the hybrid Connectionist Temporal Classification (CTC) and attention (CTC/attention) based ASR architecture. However, how to deploy hybrid CTC/attention systems for online speech recognition is still a non-trivial problem. This article describes our proposed online hybrid CTC/attention end-to-end ASR architecture, which replaces all the offline components of conventional CTC/attention ASR architecture with their corresponding streaming components. Firstly, we propose stable monotonic chunk-wise attention (sMoChA) to stream the conventional global attention, and further propose monotonic truncated attention (MTA) to simplify sMoChA and solve the training-and-decoding mismatch problem of sMoChA. Secondly, we propose truncated CTC (T-CTC) prefix score to stream CTC prefix score calculation. Thirdly, we design dynamic waiting joint decoding (DWJD) algorithm to dynamically collect the predictions of CTC and attention in an online manner. Finally, we use latency-controlled bidirectional long short-term memory (LC-BLSTM) to stream the widely-used offline bidirectional encoder network. Experiments with LibriSpeech English and HKUST Mandarin tasks demonstrate that, compared with the offline CTC/attention model, our proposed online CTC/attention model improves the real time factor in human-computer interaction services and maintains its performance with moderate degradation. To the best of our knowledge, this is the first work to provide the full-stack online solution for CTC/attention end-to-end ASR architecture.
Generative large language models (LLMs) can be a powerful tool for augmenting text annotation procedures, but their performance varies across annotation tasks due to prompt quality, text data idiosyncrasies, and conceptual difficulty. Because these challenges will persist even as LLM technology improves, we argue that any automated annotation process using an LLM must validate the LLM's performance against labels generated by humans. To this end, we outline a workflow to harness the annotation potential of LLMs in a principled, efficient way. Using GPT-4, we validate this approach by replicating 27 annotation tasks across 11 datasets from recent social science articles in high-impact journals. We find that LLM performance for text annotation is promising but highly contingent on both the dataset and the type of annotation task, which reinforces the necessity to validate on a task-by-task basis. We make available easy-to-use software designed to implement our workflow and streamline the deployment of LLMs for automated annotation.
Transformers are very powerful tools for a variety of tasks across domains, from text generation to image captioning. However, transformers require substantial amounts of training data, which is often a challenge in biomedical settings, where high quality labeled data can be challenging or expensive to obtain. This study investigates the efficacy of Compact Convolutional Transformers (CCT) for robust medical image classification with limited data, addressing a key issue faced by conventional Vision Transformers - their requirement for large datasets. A hybrid of transformers and convolutional layers, CCTs demonstrate high accuracy on modestly sized datasets. We employed a benchmark dataset of peripheral blood cell images of eight distinct cell types, each represented by approximately 2,000 low-resolution (28x28x3 pixel) samples. Despite the dataset size being smaller than those typically used with Vision Transformers, we achieved a commendable classification accuracy of 92.49% and a micro-average ROC AUC of 0.9935. The CCT also learned quickly, exceeding 80% validation accuracy after five epochs. Analysis of per-class precision, recall, F1, and ROC showed that performance was strong across cell types. Our findings underscore the robustness of CCTs, indicating their potential as a solution to data scarcity issues prevalent in biomedical imaging. We substantiate the applicability of CCTs in data-constrained areas and encourage further work on CCTs.
Product attribute-value identification (PAVI) has been studied to link products on e-commerce sites with their attribute values (e.g., <Material, Cotton>) using product text as clues. Technical demands from real-world e-commerce platforms require PAVI methods to handle unseen values, multi-attribute values, and canonicalized values, which are only partly addressed in existing extraction- and classification-based approaches. Motivated by this, we explore a generative approach to the PAVI task. We finetune a pre-trained generative model, T5, to decode a set of attribute-value pairs as a target sequence from the given product text. Since the attribute value pairs are unordered set elements, how to linearize them will matter; we, thus, explore methods of composing an attribute-value pair and ordering the pairs for the task. Experimental results confirm that our generation-based approach outperforms the existing extraction and classification-based methods on large-scale real-world datasets meant for those methods.
End-to-end speech translation (ST) is the task of translating speech signals in the source language into text in the target language. As a cross-modal task, end-to-end ST is difficult to train with limited data. Existing methods often try to transfer knowledge from machine translation (MT), but their performances are restricted by the modality gap between speech and text. In this paper, we propose Cross-modal Mixup via Optimal Transport CMOT to overcome the modality gap. We find the alignment between speech and text sequences via optimal transport and then mix up the sequences from different modalities at a token level using the alignment. Experiments on the MuST-C ST benchmark demonstrate that CMOT achieves an average BLEU of 30.0 in 8 translation directions, outperforming previous methods. Further analysis shows CMOT can adaptively find the alignment between modalities, which helps alleviate the modality gap between speech and text. Code is publicly available at https://github.com/ictnlp/CMOT.
Recent advancements in foundation models (FMs), such as GPT-4 and LLaMA, have attracted significant attention due to their exceptional performance in zero-shot learning scenarios. Similarly, in the field of visual learning, models like Grounding DINO and the Segment Anything Model (SAM) have exhibited remarkable progress in open-set detection and instance segmentation tasks. It is undeniable that these FMs will profoundly impact a wide range of real-world visual learning tasks, ushering in a new paradigm shift for developing such models. In this study, we concentrate on the remote sensing domain, where the images are notably dissimilar from those in conventional scenarios. We developed a pipeline that leverages multiple FMs to facilitate remote sensing image semantic segmentation tasks guided by text prompt, which we denote as Text2Seg. The pipeline is benchmarked on several widely-used remote sensing datasets, and we present preliminary results to demonstrate its effectiveness. Through this work, we aim to provide insights into maximizing the applicability of visual FMs in specific contexts with minimal model tuning. The code is available at https://github.com/Douglas2Code/Text2Seg.
Fact-centric question answering (QA) often requires access to multiple, heterogeneous, information sources. By jointly considering several sources like a knowledge base (KB), a text collection, and tables from the web, QA systems can enhance their answer coverage and confidence. However, existing QA benchmarks are mostly constructed with a single source of knowledge in mind. This limits capabilities of these benchmarks to fairly evaluate QA systems that can tap into more than one information repository. To bridge this gap, we release CompMix, a crowdsourced QA benchmark which naturally demands the integration of a mixture of input sources. CompMix has a total of 9,410 questions, and features several complex intents like joins and temporal conditions. Evaluation of a range of QA systems on CompMix highlights the need for further research on leveraging information from heterogeneous sources.
We introduce Noise2Music, where a series of diffusion models is trained to generate high-quality 30-second music clips from text prompts. Two types of diffusion models, a generator model, which generates an intermediate representation conditioned on text, and a cascader model, which generates high-fidelity audio conditioned on the intermediate representation and possibly the text, are trained and utilized in succession to generate high-fidelity music. We explore two options for the intermediate representation, one using a spectrogram and the other using audio with lower fidelity. We find that the generated audio is not only able to faithfully reflect key elements of the text prompt such as genre, tempo, instruments, mood, and era, but goes beyond to ground fine-grained semantics of the prompt. Pretrained large language models play a key role in this story -- they are used to generate paired text for the audio of the training set and to extract embeddings of the text prompts ingested by the diffusion models. Generated examples: https://google-research.github.io/noise2music