Xavier Amatriain

Xavier Amatriain, Ananth Sankar, Jie Bing et al.

In the past few years we have seen the meteoric appearance of dozens of foundation models of the Transformer family, all of which have memorable and sometimes funny, but not self-explanatory, names. The goal of this paper is to offer a somewhat comprehensive but simple catalog and classification of the most popular Transformer models. The paper also includes an introduction to the most important aspects and innovations in Transformer models. Our catalog will include models that are trained using self-supervised learning (e.g., BERT or GPT3) as well as those that are further trained using a human-in-the-loop (e.g. the InstructGPT model used by ChatGPT).

Mengqian Wang, Ilya Valmianski, Xavier Amatriain et al.

Medical conversations between patients and medical professionals have implicit functional sections, such as "history taking", "summarization", "education", and "care plan." In this work, we are interested in learning to automatically extract these sections. A direct approach would require collecting large amounts of expert annotations for this task, which is inherently costly due to the contextual inter-and-intra variability between these sections. This paper presents an approach that tackles the problem of learning to classify medical dialogue into functional sections without requiring a large number of annotations. Our approach combines pseudo-labeling and human-in-the-loop. First, we bootstrap using weak supervision with pseudo-labeling to generate dialogue turn-level pseudo-labels and train a transformer-based model, which is then applied to individual sentences to create noisy sentence-level labels. Second, we iteratively refine sentence-level labels using a cluster-based human-in-the-loop approach. Each iteration requires only a few dozen annotator decisions. We evaluate the results on an expert-annotated dataset of 100 dialogues and find that while our models start with 69.5% accuracy, we can iteratively improve it to 82.5%. The code used to perform all experiments described in this paper can be found here: https://github.com/curai/curai-research/tree/main/functional-sections.

Maksim Eremeev, Ilya Valmianski, Xavier Amatriain et al.

Factual correctness is often the limiting factor in practical applications of natural language generation in high-stakes domains such as healthcare. An essential requirement for maintaining factuality is the ability to deal with rare tokens. This paper focuses on rare tokens that appear in both the source and the reference sequences, and which, when missed during generation, decrease the factual correctness of the output text. For high-stake domains that are also knowledge-rich, we show how to use knowledge to (a) identify which rare tokens that appear in both source and reference are important and (b) uplift their conditional probability. We introduce the ``utilization rate'' that encodes knowledge and serves as a regularizer by maximizing the marginal probability of selected tokens. We present a study in a knowledge-rich domain of healthcare, where we tackle the problem of generating after-visit care instructions based on patient-doctor dialogues. We verify that, in our dataset, specific medical concepts with high utilization rates are underestimated by conventionally trained sequence-to-sequence models. We observe that correcting this with our approach to knowledge injection reduces the uncertainty of the model as well as improves factuality and coherence without negatively impacting fluency.

Shervin Minaee, Tomas Mikolov, Narjes Nikzad et al.

Large Language Models (LLMs) have drawn a lot of attention due to their strong performance on a wide range of natural language tasks, since the release of ChatGPT in November 2022. LLMs' ability of general-purpose language understanding and generation is acquired by training billions of model's parameters on massive amounts of text data, as predicted by scaling laws \cite{kaplan2020scaling,hoffmann2022training}. The research area of LLMs, while very recent, is evolving rapidly in many different ways. In this paper, we review some of the most prominent LLMs, including three popular LLM families (GPT, LLaMA, PaLM), and discuss their characteristics, contributions and limitations. We also give an overview of techniques developed to build, and augment LLMs. We then survey popular datasets prepared for LLM training, fine-tuning, and evaluation, review widely used LLM evaluation metrics, and compare the performance of several popular LLMs on a set of representative benchmarks. Finally, we conclude the paper by discussing open challenges and future research directions.

Raymond Li, Ilya Valmianski, Li Deng et al.

Medical entity span extraction and linking are critical steps for many healthcare NLP tasks. Most existing entity extraction methods either have a fixed vocabulary of medical entities or require span annotations. In this paper, we propose a method for linking an open set of entities that does not require any span annotations. Our method, Open Set Label Attention Transformer (OSLAT), uses the label-attention mechanism to learn candidate-entity contextualized text representations. We find that OSLAT can not only link entities but is also able to implicitly learn spans associated with entities. We evaluate OSLAT on two tasks: (1) span extraction trained without explicit span annotations, and (2) entity linking trained without span-level annotation. We test the generalizability of our method by training two separate models on two datasets with low entity overlap and comparing cross-dataset performance.

Rhys Compton, Ilya Valmianski, Li Deng et al.

We present MEDCOD, a Medically-Accurate, Emotive, Diverse, and Controllable Dialog system with a unique approach to the natural language generator module. MEDCOD has been developed and evaluated specifically for the history taking task. It integrates the advantage of a traditional modular approach to incorporate (medical) domain knowledge with modern deep learning techniques to generate flexible, human-like natural language expressions. Two key aspects of MEDCOD's natural language output are described in detail. First, the generated sentences are emotive and empathetic, similar to how a doctor would communicate to the patient. Second, the generated sentence structures and phrasings are varied and diverse while maintaining medical consistency with the desired medical concept (provided by the dialogue manager module of MEDCOD). Experimental results demonstrate the effectiveness of our approach in creating a human-like medical dialogue system. Relevant code is available at https://github.com/curai/curai-research/tree/main/MEDCOD

Varun Nair, Namit Katariya, Xavier Amatriain et al.

Medical conversation summarization is integral in capturing information gathered during interactions between patients and physicians. Summarized conversations are used to facilitate patient hand-offs between physicians, and as part of providing care in the future. Summaries, however, can be time-consuming to produce and require domain expertise. Modern pre-trained NLP models such as PEGASUS have emerged as capable alternatives to human summarization, reaching state-of-the-art performance on many summarization benchmarks. However, many downstream tasks still require at least moderately sized datasets to achieve satisfactory performance. In this work we (1) explore the effect of dataset size on transfer learning medical conversation summarization using PEGASUS and (2) evaluate various iterative labeling strategies in the low-data regime, following their success in the classification setting. We find that model performance saturates with increase in dataset size and that the various active-learning strategies evaluated all show equivalent performance consistent with simple dataset size increase. We also find that naive iterative pseudo-labeling is on-par or slightly worse than no pseudo-labeling. Our work sheds light on the successes and challenges of translating low-data regime techniques in classification to medical conversation summarization and helps guides future work in this space. Relevant code available at \url{https://github.com/curai/curai-research/tree/main/medical-summarization-ML4H-2021}.

Xavier Amatriain

Prompt design and engineering has rapidly become essential for maximizing the potential of large language models. In this paper, we introduce core concepts, advanced techniques like Chain-of-Thought and Reflection, and the principles behind building LLM-based agents. Finally, we provide a survey of tools for prompt engineers.

Bharath Chintagunta, Namit Katariya, Xavier Amatriain et al.

In medical dialogue summarization, summaries must be coherent and must capture all the medically relevant information in the dialogue. However, learning effective models for summarization require large amounts of labeled data which is especially hard to obtain. We present an algorithm to create synthetic training data with an explicit focus on capturing medically relevant information. We utilize GPT-3 as the backbone of our algorithm and scale 210 human labeled examples to yield results comparable to using 6400 human labeled examples (~30x) leveraging low-shot learning and an ensemble method. In detailed experiments, we show that this approach produces high quality training data that can further be combined with human labeled data to get summaries that are strongly preferable to those produced by models trained on human data alone both in terms of medical accuracy and coherency.

Ali Mottaghi, Prathusha K Sarma, Xavier Amatriain et al.

We study the problem of medical symptoms recognition from patient text, for the purposes of gathering pertinent information from the patient (known as history-taking). A typical patient text is often descriptive of the symptoms the patient is experiencing and a single instance of such a text can be "labeled" with multiple symptoms. This makes learning a medical symptoms recognizer challenging on account of i) the lack of availability of voluminous annotated data as well as ii) the large unknown universe of multiple symptoms that a single text can map to. Furthermore, patient text is often characterized by a long tail in the data (i.e., some labels/symptoms occur more frequently than others for e.g "fever" vs "hematochezia"). In this paper, we introduce an active learning method that leverages underlying structure of a continually refined, learned latent space to select the most informative examples to label. This enables the selection of the most informative examples that progressively increases the coverage on the universe of symptoms via the learned model, despite the long tail in data distribution.

Anirudh Joshi, Namit Katariya, Xavier Amatriain et al.

Understanding a medical conversation between a patient and a physician poses a unique natural language understanding challenge since it combines elements of standard open ended conversation with very domain specific elements that require expertise and medical knowledge. Summarization of medical conversations is a particularly important aspect of medical conversation understanding since it addresses a very real need in medical practice: capturing the most important aspects of a medical encounter so that they can be used for medical decision making and subsequent follow ups. In this paper we present a novel approach to medical conversation summarization that leverages the unique and independent local structures created when gathering a patient's medical history. Our approach is a variation of the pointer generator network where we introduce a penalty on the generator distribution, and we explicitly model negations. The model also captures important properties of medical conversations such as medical knowledge coming from standardized medical ontologies better than when those concepts are introduced explicitly. Through evaluation by doctors, we show that our approach is preferred on twice the number of summaries to the baseline pointer generator model and captures most or all of the information in 80% of the conversations making it a realistic alternative to costly manual summarization by medical experts.

Anitha Kannan, Richard Chen, Vignesh Venkataraman et al.

The COVID-19 pandemic has magnified an already existing trend of people looking for healthcare solutions online. One class of solutions are symptom checkers, which have become very popular in the context of COVID-19. Traditional symptom checkers, however, are based on manually curated expert systems that are inflexible and hard to modify, especially in a quickly changing situation like the one we are facing today. That is why all COVID-19 existing solutions are manual symptom checkers that can only estimate the probability of this disease and cannot contemplate alternative hypothesis or come up with a differential diagnosis. While machine learning offers an alternative, the lack of reliable data does not make it easy to apply to COVID-19 either. In this paper we present an approach that combines the strengths of traditional AI expert systems and novel deep learning models. In doing so we can leverage prior knowledge as well as any amount of existing data to quickly derive models that best adapt to the current state of the world and latest scientific knowledge. We use the approach to train a COVID-19 aware differential diagnosis model that can be used for medical decision support both for doctors or patients. We show that our approach is able to accurately model new incoming data about COVID-19 while still preserving accuracy on conditions that had been modeled in the past. While our approach shows evident and clear advantages for an extreme situation like the one we are currently facing, we also show that its flexibility generalizes beyond this concrete, but very important, example.

Clara H. McCreery, Namit Katariya, Anitha Kannan et al.

People increasingly search online for answers to their medical questions but the rate at which medical questions are asked online significantly exceeds the capacity of qualified people to answer them. This leaves many questions unanswered or inadequately answered. Many of these questions are not unique, and reliable identification of similar questions would enable more efficient and effective question answering schema. COVID-19 has only exacerbated this problem. Almost every government agency and healthcare organization has tried to meet the informational need of users by building online FAQs, but there is no way for people to ask their question and know if it is answered on one of these pages. While many research efforts have focused on the problem of general question similarity, these approaches do not generalize well to domains that require expert knowledge to determine semantic similarity, such as the medical domain. In this paper, we show how a double fine-tuning approach of pretraining a neural network on medical question-answer pairs followed by fine-tuning on medical question-question pairs is a particularly useful intermediate task for the ultimate goal of determining medical question similarity. While other pretraining tasks yield an accuracy below 78.7% on this task, our model achieves an accuracy of 82.6% with the same number of training examples, an accuracy of 80.0% with a much smaller training set, and an accuracy of 84.5% when the full corpus of medical question-answer data is used. We also describe a currently live system that uses the trained model to match user questions to COVID-related FAQs.

Anitha Kannan, Jason Alan Fries, Eric Kramer et al.

A third of adults in America use the Internet to diagnose medical concerns, and online symptom checkers are increasingly part of this process. These tools are powered by diagnosis models similar to clinical decision support systems, with the primary difference being the coverage of symptoms and diagnoses. To be useful to patients and physicians, these models must have high accuracy while covering a meaningful space of symptoms and diagnoses. To the best of our knowledge, this paper is the first in studying the trade-off between the coverage of the model and its performance for diagnosis. To this end, we learn diagnosis models with different coverage from EHR data. We find a 1\% drop in top-3 accuracy for every 10 diseases added to the coverage. We also observe that complexity for these models does not affect performance, with linear models performing as well as neural networks.

Sam Shleifer, Manish Chablani, Anitha Kannan et al.

Generative seq2seq dialogue systems are trained to predict the next word in dialogues that have already occurred. They can learn from large unlabeled conversation datasets, build a deeper understanding of conversational context, and generate a wide variety of responses. This flexibility comes at the cost of control, a concerning tradeoff in doctor/patient interactions. Inaccuracies, typos, or undesirable content in the training data will be reproduced by the model at inference time. We trade a small amount of labeling effort and some loss of response variety in exchange for quality control. More specifically, a pretrained language model encodes the conversational context, and we finetune a classification head to map an encoded conversational context to a response class, where each class is a noisily labeled group of interchangeable responses. Experts can update these exemplar responses over time as best practices change without retraining the classifier or invalidating old training data. Expert evaluation of 775 unseen doctor/patient conversations shows that only 12% of the discriminative model's responses are worse than the what the doctor ended up writing, compared to 18% for the generative model.

Clara McCreery, Namit Katariya, Anitha Kannan et al.

The rate at which medical questions are asked online far exceeds the capacity of qualified people to answer them, and many of these questions are not unique. Identifying same-question pairs could enable questions to be answered more effectively. While many research efforts have focused on the problem of general question similarity for non-medical applications, these approaches do not generalize well to the medical domain, where medical expertise is often required to determine semantic similarity. In this paper, we show how a semi-supervised approach of pre-training a neural network on medical question-answer pairs is a particularly useful intermediate task for the ultimate goal of determining medical question similarity. While other pre-training tasks yield an accuracy below 78.7% on this task, our model achieves an accuracy of 82.6% with the same number of training examples, and an accuracy of 80.0% with a much smaller training set.

Viraj Prabhu, Anitha Kannan, Geoffrey J. Tso et al.

Machine-learned diagnosis models have shown promise as medical aides but are trained under a closed-set assumption, i.e. that models will only encounter conditions on which they have been trained. However, it is practically infeasible to obtain sufficient training data for every human condition, and once deployed such models will invariably face previously unseen conditions. We frame machine-learned diagnosis as an open-set learning problem, and study how state-of-the-art approaches compare. Further, we extend our study to a setting where training data is distributed across several healthcare sites that do not allow data pooling, and experiment with different strategies of building open-set diagnostic ensembles. Across both settings, we observe consistent gains from explicitly modeling unseen conditions, but find the optimal training strategy to vary across settings.

Sam Shleifer, Manish Chablani, Namit Katariya et al.

Generative seq2seq dialogue systems are trained to predict the next word in dialogues that have already occurred. They can learn from large unlabeled conversation datasets, build a deep understanding of conversational context, and generate a wide variety of responses. This flexibility comes at the cost of control. Undesirable responses in the training data will be reproduced by the model at inference time, and longer generations often don't make sense. Instead of generating responses one word at a time, we train a classifier to choose from a predefined list of full responses. The classifier is trained on (conversation context, response class) pairs, where each response class is a noisily labeled group of interchangeable responses. At inference, we generate the exemplar response associated with the predicted response class. Experts can edit and improve these exemplar responses over time without retraining the classifier or invalidating old training data. Human evaluation of 775 unseen doctor/patient conversations shows that this tradeoff improves responses. Only 12% of our discriminative approach's responses are worse than the doctor's response in the same conversational context, compared to 18% for the generative model. A discriminative model trained without any manual labeling of response classes achieves equal performance to the generative model.

Viraj Prabhu, Anitha Kannan, Murali Ravuri et al.

We consider the problem of image classification for the purpose of aiding doctors in dermatological diagnosis. Dermatological diagnosis poses two major challenges for standard off-the-shelf techniques: First, the data distribution is typically extremely long tailed. Second, intra-class variability is often large. To address the first issue, we formulate the problem as low-shot learning, where once deployed, a base classifier must rapidly generalize to diagnose novel conditions given very few labeled examples. To model diverse classes effectively, we propose Prototypical Clustering Networks (PCN), an extension to Prototypical Networks that learns a mixture of prototypes for each class. Prototypes are initialized for each class via clustering and refined via an online update scheme. Classification is performed by measuring similarity to a weighted combination of prototypes within a class, where the weights are the inferred cluster responsibilities. We demonstrate the strengths of our approach in effective diagnosis on a realistic dataset of dermatological conditions.

Murali Ravuri, Anitha Kannan, Geoffrey J. Tso et al.

Expert diagnostic support systems have been extensively studied. The practical applications of these systems in real-world scenarios have been somewhat limited due to well-understood shortcomings, such as lack of extensibility. More recently, machine-learned models for medical diagnosis have gained momentum, since they can learn and generalize patterns found in very large datasets like electronic health records. These models also have shortcomings - in particular, there is no easy way to incorporate prior knowledge from existing literature or experts. In this paper, we present a method to merge both approaches by using expert systems as generative models that create simulated data on which models can be learned. We demonstrate that such a learned model not only preserves the original properties of the expert systems but also addresses some of their limitations. Furthermore, we show how this approach can also be used as the starting point to combine expert knowledge with knowledge extracted from other data sources, such as electronic health records.