The bias-variance tradeoff is the idea that learning methods need to balance model complexity with data size to minimize both under-fitting and over-fitting. Recent empirical work and theoretical analysis with over-parameterized neural networks challenges the classic bias-variance trade-off notion suggesting that no such trade-off holds: as the width of the network grows, bias monotonically decreases while variance initially increases followed by a decrease. In this work, we first provide a variance decomposition-based justification criteria to examine whether large pretrained neural models in a fine-tuning setting are generalizable enough to have low bias and variance. We then perform theoretical and empirical analysis using ensemble methods explicitly designed to decrease variance due to optimization. This results in essentially a two-stage fine-tuning algorithm that first ratchets down bias and variance iteratively, and then uses a selected fixed-bias model to further reduce variance due to optimization by ensembling. We also analyze the nature of variance change with the ensemble size in low- and high-resource classes. Empirical results show that this two-stage method obtains strong results on SuperGLUE tasks and clinical information extraction tasks. Code and settings are available: https://github.com/christa60/bias-var-fine-tuning-plms.git

Generative artificial intelligence (AI) is a promising direction for augmenting clinical diagnostic decision support and reducing diagnostic errors, a leading contributor to medical errors. To further the development of clinical AI systems, the Diagnostic Reasoning Benchmark (DR.BENCH) was introduced as a comprehensive generative AI framework, comprised of six tasks representing key components in clinical reasoning. We present a comparative analysis of in-domain versus out-of-domain language models as well as multi-task versus single task training with a focus on the problem summarization task in DR.BENCH. We demonstrate that a multi-task, clinically-trained language model outperforms its general domain counterpart by a large margin, establishing a new state-of-the-art performance, with a ROUGE-L score of 28.55. This research underscores the value of domain-specific training for optimizing clinical diagnostic reasoning tasks.

Text in electronic health records is organized into sections, and classifying those sections into section categories is useful for downstream tasks. In this work, we attempt to improve the transferability of section classification models by combining the dataset-specific knowledge in supervised learning models with the world knowledge inside large language models (LLMs). Surprisingly, we find that zero-shot LLMs out-perform supervised BERT-based models applied to out-of-domain data. We also find that their strengths are synergistic, so that a simple ensemble technique leads to additional performance gains.

Understanding temporal relationships in text from electronic health records can be valuable for many important downstream clinical applications. Since Clinical TempEval 2017, there has been little work on end-to-end systems for temporal relation extraction, with most work focused on the setting where gold standard events and time expressions are given. In this work, we make use of a novel multi-headed attention mechanism on top of a pre-trained transformer encoder to allow the learning process to attend to multiple aspects of the contextualized embeddings. Our system achieves state of the art results on the THYME corpus by a wide margin, in both the in-domain and cross-domain settings.

The BioNLP Workshop 2023 initiated the launch of a shared task on Problem List Summarization (ProbSum) in January 2023. The aim of this shared task is to attract future research efforts in building NLP models for real-world diagnostic decision support applications, where a system generating relevant and accurate diagnoses will augment the healthcare providers’ decision-making process and improve the quality of care for patients. The goal for participants is to develop models that generated a list of diagnoses and problems using input from the daily care notes collected from the hospitalization of critically ill patients. Eight teams submitted their final systems to the shared task leaderboard. In this paper, we describe the tasks, datasets, evaluation metrics, and baseline systems. Additionally, the techniques and results of the evaluation of the different approaches tried by the participating teams are summarized.

In this paper, we investigate ensemble methods for fine-tuning transformer-based pretrained models for clinical natural language processing tasks, specifically temporal relation extraction from the clinical narrative. Our experimental results on the THYME data show that ensembling as a fine-tuning strategy can further boost model performance over single learners optimized for hyperparameters. Dynamic snapshot ensembling is particularly beneficial as it fine-tunes a wide array of parameters and results in a 2.8% absolute improvement in F1 over the base single learner.

Sequence-to-sequence models are appealing because they allow both encoder and decoder to be shared across many tasks by formulating those tasks as text-to-text problems. Despite recently reported successes of such models, we find that engineering input/output representations for such text-to-text models is challenging. On the Clinical TempEval 2016 relation extraction task, the most natural choice of output representations, where relations are spelled out in simple predicate logic statements, did not lead to good performance. We explore a variety of input/output representations, with the most successful prompting one event at a time, and achieving results competitive with standard pairwise temporal relation extraction systems.

Applying methods in natural language processing on electronic health records (EHR) data has attracted rising interests. Existing corpus and annotation focus on modeling textual features and relation prediction. However, there are a paucity of annotated corpus built to model clinical diagnostic thinking, a processing involving text understanding, domain knowledge abstraction and reasoning. In this work, we introduce a hierarchical annotation schema with three stages to address clinical text understanding, clinical reasoning and summarization. We create an annotated corpus based on a large collection of publicly available daily progress notes, a type of EHR that is time-sensitive, problem-oriented, and well-documented by the format of Subjective, Objective, Assessment and Plan (SOAP). We also define a new suite of tasks, Progress Note Understanding, with three tasks utilizing the three annotation stages. This new suite aims at training and evaluating future NLP models for clinical text understanding, clinical knowledge representation, inference and summarization.

Automatically summarizing patients’ main problems from daily progress notes using natural language processing methods helps to battle against information and cognitive overload in hospital settings and potentially assists providers with computerized diagnostic decision support. Problem list summarization requires a model to understand, abstract, and generate clinical documentation. In this work, we propose a new NLP task that aims to generate a list of problems in a patient’s daily care plan using input from the provider’s progress notes during hospitalization. We investigate the performance of T5 and BART, two state-of-the-art seq2seq transformer architectures, in solving this problem. We provide a corpus built on top of progress notes from publicly available electronic health record progress notes in the Medical Information Mart for Intensive Care (MIMIC)-III. T5 and BART are trained on general domain text, and we experiment with a data augmentation method and a domain adaptation pre-training method to increase exposure to medical vocabulary and knowledge. Evaluation methods include ROUGE, BERTScore, cosine similarity on sentence embedding, and F-score on medical concepts. Results show that T5 with domain adaptive pre-training achieves significant performance gains compared to a rule-based system and general domain pre-trained language models, indicating a promising direction for tackling the problem summarization task.

This paper presents the Source-Free Domain Adaptation shared task held within SemEval-2021. The aim of the task was to explore adaptation of machine-learning models in the face of data sharing constraints. Specifically, we consider the scenario where annotations exist for a domain but cannot be shared. Instead, participants are provided with models trained on that (source) data. Participants also receive some labeled data from a new (development) domain on which to explore domain adaptation algorithms. Participants are then tested on data representing a new (target) domain. We explored this scenario with two different semantic tasks: negation detection (a text classification task) and time expression recognition (a sequence tagging task).

Advances in transfer learning and domain adaptation have raised hopes that once-challenging NLP tasks are ready to be put to use for sophisticated information extraction needs. In this work, we describe an effort to do just that – combining state-of-the-art neural methods for negation detection, document time relation extraction, and aspectual link prediction, with the eventual goal of extracting drug timelines from electronic health record text. We train on the THYME colon cancer corpus and test on both the THYME brain cancer corpus and an internal corpus, and show that performance of the combined systems is unacceptable despite good performance of individual systems. Although domain adaptation shows improvements on each individual system, the model selection problem is a barrier to improving overall pipeline performance.

This article describes a simple PCFG induction model with a fixed category domain that predicts a large majority of attested constituent boundaries, and predicts labels consistent with nearly half of attested constituent labels on a standard evaluation data set of child-directed speech. The article then explores the idea that the difference between simple grammars exhibited by child learners and fully recursive grammars exhibited by adult learners may be an effect of increasing working memory capacity, where the shallow grammars are constrained images of the recursive grammars. An implementation of these memory bounds as limits on center embedding in a depth-specific transform of a recursive grammar yields a significant improvement over an equivalent but unbounded baseline, suggesting that this arrangement may indeed confer a learning advantage.

Transformer-based neural language models have led to breakthroughs for a variety of natural language processing (NLP) tasks. However, most models are pretrained on general domain data. We propose a methodology to produce a model focused on the clinical domain: continued pretraining of a model with a broad representation of biomedical terminology (PubMedBERT) on a clinical corpus along with a novel entity-centric masking strategy to infuse domain knowledge in the learning process. We show that such a model achieves superior results on clinical extraction tasks by comparing our entity-centric masking strategy with classic random masking on three clinical NLP tasks: cross-domain negation detection, document time relation (DocTimeRel) classification, and temporal relation extraction. We also evaluate our models on the PubMedQA dataset to measure the models’ performance on a non-entity-centric task in the biomedical domain. The language addressed in this work is English.

Reducing rates of early hospital readmission has been recognized and identified as a key to improve quality of care and reduce costs. There are a number of risk factors that have been hypothesized to be important for understanding re-admission risk, including such factors as problems with substance abuse, ability to maintain work, relations with family. In this work, we develop Roberta-based models to predict the sentiment of sentences describing readmission risk factors in discharge summaries of patients with psychosis. We improve substantially on previous results by a scheme that shares information across risk factors while also allowing the model to learn risk factor-specific information.

We present work on extraction of radiotherapy treatment information from the clinical narrative in the electronic medical records. Radiotherapy is a central component of the treatment of most solid cancers. Its details are described in non-standardized fashions using jargon not found in other medical specialties, complicating the already difficult task of manual data extraction. We examine the performance of several state-of-the-art neural methods for relation extraction of radiotherapy treatment details, with a goal of automating detailed information extraction. The neural systems perform at 0.82-0.88 macro-average F1, which approximates or in some cases exceeds the inter-annotator agreement. To the best of our knowledge, this is the first effort to develop models for radiotherapy relation extraction and one of the few efforts for relation extraction to describe cancer treatment in general.

We present refinements over existing temporal relation annotations in the Electronic Medical Record clinical narrative. We refined the THYME corpus annotations to more faithfully represent nuanced temporality and nuanced temporal-coreferential relations. The main contributions are in re-defining CONTAINS and OVERLAP relations into CONTAINS, CONTAINS-SUBEVENT, OVERLAP and NOTED-ON. We demonstrate that these refinements lead to substantial gains in learnability for state-of-the-art transformer models as compared to previously reported results on the original THYME corpus. We thus establish a baseline for the automatic extraction of these refined temporal relations. Although our study is done on clinical narrative, we believe it addresses far-reaching challenges that are corpus- and domain- agnostic.

Electronic consult (eConsult) systems allow specialists more flexibility to respond to referrals more efficiently, thereby increasing access in under-resourced healthcare settings like safety net systems. Understanding the usage patterns of eConsult system is an important part of improving specialist efficiency. In this work, we develop and apply classifiers to a dataset of eConsult questions from primary care providers to specialists, classifying the messages for how they were triaged by the specialist office, and the underlying type of clinical question posed by the primary care provider. We show that pre-trained transformer models are strong baselines, with improving performance from domain-specific training and shared representations.

Recently BERT has achieved a state-of-the-art performance in temporal relation extraction from clinical Electronic Medical Records text. However, the current approach is inefficient as it requires multiple passes through each input sequence. We extend a recently-proposed one-pass model for relation classification to a one-pass model for relation extraction. We augment this framework by introducing global embeddings to help with long-distance relation inference, and by multi-task learning to increase model performance and generalizability. Our proposed model produces results on par with the state-of-the-art in temporal relation extraction on the THYME corpus and is much “greener” in computational cost.

In this paper we describe an evaluation of the potential of classical information extraction methods to extract drug-related attributes, including adverse drug events, and compare to more recently developed neural methods. We use the 2018 N2C2 shared task data as our gold standard data set for training. We train support vector machine classifiers to detect drug and drug attribute spans, and pair these detected entities as training instances for an SVM relation classifier, with both systems using standard features. We compare to baseline neural methods that use standard contextualized embedding representations for entity and relation extraction. The SVM-based system and a neural system obtain comparable results, with the SVM system doing better on concepts and the neural system performing better on relation extraction tasks. The neural system obtains surprisingly strong results compared to the system based on years of research in developing features for information extraction.

Classic methods for clinical temporal relation extraction focus on relational candidates within a sentence. On the other hand, break-through Bidirectional Encoder Representations from Transformers (BERT) are trained on large quantities of arbitrary spans of contiguous text instead of sentences. In this study, we aim to build a sentence-agnostic framework for the task of CONTAINS temporal relation extraction. We establish a new state-of-the-art result for the task, 0.684F for in-domain (0.055-point improvement) and 0.565F for cross-domain (0.018-point improvement), by fine-tuning BERT and pre-training domain-specific BERT models on sentence-agnostic temporal relation instances with WordPiece-compatible encodings, and augmenting the labeled data with automatically generated “silver” instances.

A large percentage of medical information is in unstructured text format in electronic medical record systems. Manual extraction of information from clinical notes is extremely time consuming. Natural language processing has been widely used in recent years for automatic information extraction from medical texts. However, algorithms trained on data from a single healthcare provider are not generalizable and error-prone due to the heterogeneity and uniqueness of medical documents. We develop a two-stage federated natural language processing method that enables utilization of clinical notes from different hospitals or clinics without moving the data, and demonstrate its performance using obesity and comorbities phenotyping as medical task. This approach not only improves the quality of a specific clinical task but also facilitates knowledge progression in the whole healthcare system, which is an essential part of learning health system. To the best of our knowledge, this is the first application of federated machine learning in clinical NLP.

Unsupervised PCFG inducers hypothesize sets of compact context-free rules as explanations for sentences. PCFG induction not only provides tools for low-resource languages, but also plays an important role in modeling language acquisition (Bannard et al., 2009; Abend et al. 2017). However, current PCFG induction models, using word tokens as input, are unable to incorporate semantics and morphology into induction, and may encounter issues of sparse vocabulary when facing morphologically rich languages. This paper describes a neural PCFG inducer which employs context embeddings (Peters et al., 2018) in a normalizing flow model (Dinh et al., 2015) to extend PCFG induction to use semantic and morphological information. Linguistically motivated sparsity and categorical distance constraints are imposed on the inducer as regularization. Experiments show that the PCFG induction model with normalizing flow produces grammars with state-of-the-art accuracy on a variety of different languages. Ablation further shows a positive effect of normalizing flow, context embeddings and proposed regularizers.

This paper discusses a cross-document coreference annotation schema that was developed to further automatic extraction of timelines in the clinical domain. Lexical senses and coreference choices are determined largely by context, but cross-document work requires reasoning across contexts that are not necessarily coherent. We found that an annotation approach that relies less on context-guided annotator intuitions and more on schematic rules was most effective in creating meaningful and consistent cross-document relations.

Unsupervised domain adaptation (UDA) is the task of training a statistical model on labeled data from a source domain to achieve better performance on data from a target domain, with access to only unlabeled data in the target domain. Existing state-of-the-art UDA approaches use neural networks to learn representations that are trained to predict the values of subset of important features called “pivot features” on combined data from the source and target domains. In this work, we show that it is possible to improve on existing neural domain adaptation algorithms by 1) jointly training the representation learner with the task learner; and 2) removing the need for heuristically-selected “pivot features.” Our results show competitive performance with a simpler model.

There has been recent interest in applying cognitively- or empirically-motivated bounds on recursion depth to limit the search space of grammar induction models (Ponvert et al., 2011; Noji and Johnson, 2016; Shain et al., 2016). This work extends this depth-bounding approach to probabilistic context-free grammar induction (DB-PCFG), which has a smaller parameter space than hierarchical sequence models, and therefore more fully exploits the space reductions of depth-bounding. Results for this model on grammar acquisition from transcribed child-directed speech and newswire text exceed or are competitive with those of other models when evaluated on parse accuracy. Moreover, grammars acquired from this model demonstrate a consistent use of category labels, something which has not been demonstrated by other acquisition models.

There have been several recent attempts to improve the accuracy of grammar induction systems by bounding the recursive complexity of the induction model. Modern depth-bounded grammar inducers have been shown to be more accurate than early unbounded PCFG inducers, but this technique has never been compared against unbounded induction within the same system, in part because most previous depth-bounding models are built around sequence models, the complexity of which grows exponentially with the maximum allowed depth. The present work instead applies depth bounds within a chart-based Bayesian PCFG inducer, where bounding can be switched on and off, and then samples trees with or without bounding. Results show that depth-bounding is indeed significantly effective in limiting the search space of the inducer and thereby increasing accuracy of resulting parsing model, independent of the contribution of modern Bayesian induction techniques. Moreover, parsing results on English, Chinese and German show that this bounded model is able to produce parse trees more accurately than or competitively with state-of-the-art constituency grammar induction models.

Automatic identification of spurious instances (those with potentially wrong labels in datasets) can improve the quality of existing language resources, especially when annotations are obtained through crowdsourcing or automatically generated based on coded rankings. In this paper, we present effective approaches inspired by queueing theory and psychology of learning to automatically identify spurious instances in datasets. Our approaches discriminate instances based on their “difficulty to learn,” determined by a downstream learner. Our methods can be applied to any dataset assuming the existence of a neural network model for the target task of the dataset. Our best approach outperforms competing state-of-the-art baselines and has a MAP of 0.85 and 0.22 in identifying spurious instances in synthetic and carefully-crowdsourced real-world datasets respectively.

Neural network models are oftentimes restricted by limited labeled instances and resort to advanced architectures and features for cutting edge performance. We propose to build a recurrent neural network with multiple semantically heterogeneous embeddings within a self-training framework. Our framework makes use of labeled, unlabeled, and social media data, operates on basic features, and is scalable and generalizable. With this method, we establish the state-of-the-art result for both in- and cross-domain for a clinical temporal relation extraction task.

Mining electronic health records for patients who satisfy a set of predefined criteria is known in medical informatics as phenotyping. Phenotyping has numerous applications such as outcome prediction, clinical trial recruitment, and retrospective studies. Supervised machine learning for phenotyping typically relies on sparse patient representations such as bag-of-words. We consider an alternative that involves learning patient representations. We develop a neural network model for learning patient representations and show that the learned representations are general enough to obtain state-of-the-art performance on a standard comorbidity detection task.

We present a novel approach for training artificial neural networks. Our approach is inspired by broad evidence in psychology that shows human learners can learn efficiently and effectively by increasing intervals of time between subsequent reviews of previously learned materials (spaced repetition). We investigate the analogy between training neural models and findings in psychology about human memory model and develop an efficient and effective algorithm to train neural models. The core part of our algorithm is a cognitively-motivated scheduler according to which training instances and their “reviews” are spaced over time. Our algorithm uses only 34-50% of data per epoch, is 2.9-4.8 times faster than standard training, and outperforms competing state-of-the-art baselines. Our code is available at scholar.harvard.edu/hadi/RbF/.

We experiment with neural architectures for temporal relation extraction and establish a new state-of-the-art for several scenarios. We find that neural models with only tokens as input outperform state-of-the-art hand-engineered feature-based models, that convolutional neural networks outperform LSTM models, and that encoding relation arguments with XML tags outperforms a traditional position-based encoding.

Detecting negated concepts in clinical texts is an important part of NLP information extraction systems. However, generalizability of negation systems is lacking, as cross-domain experiments suffer dramatic performance losses. We examine the performance of multiple unsupervised domain adaptation algorithms on clinical negation detection, finding only modest gains that fall well short of in-domain performance.

Token sequences are often used as the input for Convolutional Neural Networks (CNNs) in natural language processing. However, they might not be an ideal representation for time expressions, which are long, highly varied, and semantically complex. We describe a method for representing time expressions with single pseudo-tokens for CNNs. With this method, we establish a new state-of-the-art result for a clinical temporal relation extraction task.

This paper presents a new memory-bounded left-corner parsing model for unsupervised raw-text syntax induction, using unsupervised hierarchical hidden Markov models (UHHMM). We deploy this algorithm to shed light on the extent to which human language learners can discover hierarchical syntax through distributional statistics alone, by modeling two widely-accepted features of human language acquisition and sentence processing that have not been simultaneously modeled by any existing grammar induction algorithm: (1) a left-corner parsing strategy and (2) limited working memory capacity. To model realistic input to human language learners, we evaluate our system on a corpus of child-directed speech rather than typical newswire corpora. Results beat or closely match those of three competing systems.

This article discusses the requirements of a formal specification for the annotation of temporal information in clinical narratives. We discuss the implementation and extension of ISO-TimeML for annotating a corpus of clinical notes, known as the THYME corpus. To reflect the information task and the heavily inference-based reasoning demands in the domain, a new annotation guideline has been developed, “the THYME Guidelines to ISO-TimeML (THYME-TimeML)”. To clarify what relations merit annotation, we distinguish between linguistically-derived and inferentially-derived temporal orderings in the text. We also apply a top performing TempEval 2013 system against this new resource to measure the difficulty of adapting systems to the clinical domain. The corpus is available to the community and has been proposed for use in a SemEval 2015 task.