Word error rate (WER) and character error rate (CER) are standard metrics in Speech Recognition (ASR), but one problem has always been alternative spellings: If one's system transcribes adviser whereas the ground truth has advisor, this will count as an error even though the two spellings really represent the same word. Japanese is notorious for ``lacking orthography'': most words can be spelled in multiple ways, presenting a problem for accurate ASR evaluation. In this paper we propose a new lenient evaluation metric as a more defensible CER measure for Japanese ASR. We create a lattice of plausible respellings of the reference transcription, using a combination of lexical resources, a Japanese text-processing system, and a neural machine translation model for reconstructing kanji from hiragana or katakana. In a manual evaluation, raters rated 95.4% of the proposed spelling variants as plausible. ASR results show that our method, which does not penalize the system for choosing a valid alternate spelling of a word, affords a 2.4%-3.1% absolute reduction in CER depending on the task.
Experimenters often collect baseline data to study heterogeneity. I propose the first valid confidence intervals for the VCATE, the treatment effect variance explained by observables. Conventional approaches yield incorrect coverage when the VCATE is zero. As a result, practitioners could be prone to detect heterogeneity even when none exists. The reason why coverage worsens at the boundary is that all efficient estimators have a locally-degenerate influence function and may not be asymptotically normal. I solve the problem for a broad class of multistep estimators with a predictive first stage. My confidence intervals account for higher-order terms in the limiting distribution and are fast to compute. I also find new connections between the VCATE and the problem of deciding whom to treat. The gains of targeting treatment are (sharply) bounded by half the square root of the VCATE. Finally, I document excellent performance in simulation and reanalyze an experiment from Malawi.
The recent development of generative large language models (LLMs) poses new challenges for model evaluation that the research community and industry have been grappling with. While the versatile capabilities of these models ignite much excitement, they also inevitably make a leap toward homogenization: powering a wide range of applications with a single, often referred to as ``general-purpose'', model. In this position paper, we argue that model evaluation practices must take on a critical task to cope with the challenges and responsibilities brought by this homogenization: providing valid assessments for whether and how much human needs in diverse downstream use cases can be satisfied by the given model (\textit{socio-technical gap}). By drawing on lessons about improving research realism from the social sciences, human-computer interaction (HCI), and the interdisciplinary field of explainable AI (XAI), we urge the community to develop evaluation methods based on real-world contexts and human requirements, and embrace diverse evaluation methods with an acknowledgment of trade-offs between realisms and pragmatic costs to conduct the evaluation. By mapping HCI and current NLG evaluation methods, we identify opportunities for evaluation methods for LLMs to narrow the socio-technical gap and pose open questions.
Structured text extraction is one of the most valuable and challenging application directions in the field of Document AI. However, the scenarios of past benchmarks are limited, and the corresponding evaluation protocols usually focus on the submodules of the structured text extraction scheme. In order to eliminate these problems, we organized the ICDAR 2023 competition on Structured text extraction from Visually-Rich Document images (SVRD). We set up two tracks for SVRD including Track 1: HUST-CELL and Track 2: Baidu-FEST, where HUST-CELL aims to evaluate the end-to-end performance of Complex Entity Linking and Labeling, and Baidu-FEST focuses on evaluating the performance and generalization of Zero-shot / Few-shot Structured Text extraction from an end-to-end perspective. Compared to the current document benchmarks, our two tracks of competition benchmark enriches the scenarios greatly and contains more than 50 types of visually-rich document images (mainly from the actual enterprise applications). The competition opened on 30th December, 2022 and closed on 24th March, 2023. There are 35 participants and 91 valid submissions received for Track 1, and 15 participants and 26 valid submissions received for Track 2. In this report we will presents the motivation, competition datasets, task definition, evaluation protocol, and submission summaries. According to the performance of the submissions, we believe there is still a large gap on the expected information extraction performance for complex and zero-shot scenarios. It is hoped that this competition will attract many researchers in the field of CV and NLP, and bring some new thoughts to the field of Document AI.
Knowledge graph embedding models (KGEMs) have gained considerable traction in recent years. These models learn a vector representation of knowledge graph entities and relations, a.k.a. knowledge graph embeddings (KGEs). Learning versatile KGEs is desirable as it makes them useful for a broad range of tasks. However, KGEMs are usually trained for a specific task, which makes their embeddings task-dependent. In parallel, the widespread assumption that KGEMs actually create a semantic representation of the underlying entities and relations (e.g., project similar entities closer than dissimilar ones) has been challenged. In this work, we design heuristics for generating protographs -- small, modified versions of a KG that leverage RDF/S information. The learnt protograph-based embeddings are meant to encapsulate the semantics of a KG, and can be leveraged in learning KGEs that, in turn, also better capture semantics. Extensive experiments on various evaluation benchmarks demonstrate the soundness of this approach, which we call Modular and Agnostic SCHema-based Integration of protograph Embeddings (MASCHInE). In particular, MASCHInE helps produce more versatile KGEs that yield substantially better performance for entity clustering and node classification tasks. For link prediction, using MASCHinE substantially increases the number of semantically valid predictions with equivalent rank-based performance.
In this paper, we investigate the theoretical properties of stochastic gradient descent (SGD) for statistical inference in the context of nonconvex optimization problems, which have been relatively unexplored compared to convex settings. Our study is the first to establish provable inferential procedures using the SGD estimator for general nonconvex objective functions, which may contain multiple local minima. We propose two novel online inferential procedures that combine SGD and the multiplier bootstrap technique. The first procedure employs a consistent covariance matrix estimator, and we establish its error convergence rate. The second procedure approximates the limit distribution using bootstrap SGD estimators, yielding asymptotically valid bootstrap confidence intervals. We validate the effectiveness of both approaches through numerical experiments. Furthermore, our analysis yields an intermediate result: the in-expectation error convergence rate for the original SGD estimator in nonconvex settings, which is comparable to existing results for convex problems. We believe this novel finding holds independent interest and enriches the literature on optimization and statistical inference.
Conformal prediction is a theoretically grounded framework for constructing predictive intervals. We study conformal prediction with missing values in the covariates -- a setting that brings new challenges to uncertainty quantification. We first show that the marginal coverage guarantee of conformal prediction holds on imputed data for any missingness distribution and almost all imputation functions. However, we emphasize that the average coverage varies depending on the pattern of missing values: conformal methods tend to construct prediction intervals that under-cover the response conditionally to some missing patterns. This motivates our novel generalized conformalized quantile regression framework, missing data augmentation, which yields prediction intervals that are valid conditionally to the patterns of missing values, despite their exponential number. We then show that a universally consistent quantile regression algorithm trained on the imputed data is Bayes optimal for the pinball risk, thus achieving valid coverage conditionally to any given data point. Moreover, we examine the case of a linear model, which demonstrates the importance of our proposal in overcoming the heteroskedasticity induced by missing values. Using synthetic and data from critical care, we corroborate our theory and report improved performance of our methods.
Equivalence classes of DAGs (represented by CPDAGs) may be too large to provide useful causal information. Here, we address incorporating tiered background knowledge yielding restricted equivalence classes represented by 'tiered MPDAGs'. Tiered knowledge leads to considerable gains in informativeness and computational efficiency: We show that construction of tiered MPDAGs only requires application of Meek's 1st rule, and that tiered MPDAGs (unlike general MPDAGs) are chain graphs with chordal components. This entails simplifications e.g. of determining valid adjustment sets for causal effect estimation. Further, we characterise when one tiered ordering is more informative than another, providing insights into useful aspects of background knowledge.
Task-oriented dialogue is difficult in part because it involves understanding user intent, collecting information from the user, executing API calls, and generating helpful and fluent responses. However, for complex tasks one must also correctly do all of these things over multiple steps, and in a specific order. While large pre-trained language models can be fine-tuned end-to-end to create multi-step task-oriented dialogue agents that generate fluent text, our experiments confirm that this approach alone cannot reliably perform new multi-step tasks that are unseen during training. To address these limitations, we augment the dialogue contexts given to \textmd{text2text} transformers with known \textit{valid workflow names} and \textit{action plans}. Action plans consist of sequences of actions required to accomplish a task, and are encoded as simple sequences of keywords (e.g. verify-identity, pull-up-account, reset-password, etc.). We perform extensive experiments on the Action-Based Conversations Dataset (ABCD) with T5-small, base and large models, and show that such models: a) are able to more readily generalize to unseen workflows by following the provided plan, and b) are able to generalize to executing unseen actions if they are provided in the plan. In contrast, models are unable to fully accomplish new multi-step tasks when they are not provided action plan information, even when given new valid workflow names.
It is known that the multiplication of an $N \times M$ matrix with an $M \times P$ matrix can be performed using fewer multiplications than what the naive $NMP$ approach suggests. The most famous instance of this is Strassen's algorithm for multiplying two $2\times 2$ matrices in 7 instead of 8 multiplications. This gives rise to the constraint satisfaction problem of fast matrix multiplication, where a set of $R < NMP$ multiplication terms must be chosen and combined such that they satisfy correctness constraints on the output matrix. Despite its highly combinatorial nature, this problem has not been exhaustively examined from that perspective, as evidenced for example by the recent deep reinforcement learning approach of AlphaTensor. In this work, we propose a simple yet novel Constraint Programming approach to find non-commutative algorithms for fast matrix multiplication or provide proof of infeasibility otherwise. We propose a set of symmetry-breaking constraints and valid inequalities that are particularly helpful in proving infeasibility. On the feasible side, we find that exploiting solver performance variability in conjunction with a sparsity-based problem decomposition enables finding solutions for larger (feasible) instances of fast matrix multiplication. Our experimental results using CP Optimizer demonstrate that we can find fast matrix multiplication algorithms for matrices up to $3\times 3$ in a short amount of time.
Recent advances to combine structured regression models and deep neural networks for better interpretability, more expressiveness, and statistically valid uncertainty quantification demonstrate the versatility of semi-structured neural networks (SSNs). We show that techniques to properly identify the contributions of the different model components in SSNs, however, lead to suboptimal network estimation, slower convergence, and degenerated or erroneous predictions. In order to solve these problems while preserving favorable model properties, we propose a non-invasive post-hoc orthogonalization (PHO) that guarantees identifiability of model components and provides better estimation and prediction quality. Our theoretical findings are supported by numerical experiments, a benchmark comparison as well as a real-world application to COVID-19 infections.
Existing object pose estimation methods commonly require a one-to-one point matching step that forces them to be separated into two consecutive stages: visual correspondence detection (e.g., by matching feature descriptors as part of a perception front-end) followed by geometric alignment (e.g., by optimizing a robust estimation objective for pointcloud registration or perspective-n-point). Instead, we propose a matching-free probabilistic formulation with two main benefits: i) it enables unified and concurrent optimization of both visual correspondence and geometric alignment, and ii) it can represent different plausible modes of the entire distribution of likely poses. This in turn allows for a more graceful treatment of geometric perception scenarios where establishing one-to-one matches between points is conceptually ill-defined, such as textureless, symmetrical and/or occluded objects and scenes where the correct pose is uncertain or there are multiple equally valid solutions.
Argument mining involves multiple sub-tasks that automatically identify argumentative elements, such as claim detection, evidence extraction, stance classification, etc. However, each subtask alone is insufficient for a thorough understanding of the argumentative structure and reasoning process. To learn a complete view of an argument essay and capture the interdependence among argumentative components, we need to know what opinions people hold (i.e., claims), why those opinions are valid (i.e., supporting evidence), which source the evidence comes from (i.e., evidence type), and how those claims react to the debating topic (i.e., stance). In this work, we for the first time propose a challenging argument quadruplet extraction task (AQE), which can provide an all-in-one extraction of four argumentative components, i.e., claims, evidence, evidence types, and stances. To support this task, we construct a large-scale and challenging dataset. However, there is no existing method that can solve the argument quadruplet extraction. To fill this gap, we propose a novel quad-tagging augmented generative approach, which leverages a quadruplet tagging module to augment the training of the generative framework. The experimental results on our dataset demonstrate the empirical superiority of our proposed approach over several strong baselines.
Heterogeneous Graph Neural Networks (HGNNs) have gained significant popularity in various heterogeneous graph learning tasks. However, most existing HGNNs rely on spatial domain-based methods to aggregate information, i.e., manually selected meta-paths or some heuristic modules, lacking theoretical guarantees. Furthermore, these methods cannot learn arbitrary valid heterogeneous graph filters within the spectral domain, which have limited expressiveness. To tackle these issues, we present a positive spectral heterogeneous graph convolution via positive noncommutative polynomials. Then, using this convolution, we propose PSHGCN, a novel Positive Spectral Heterogeneous Graph Convolutional Network. PSHGCN offers a simple yet effective method for learning valid heterogeneous graph filters. Moreover, we demonstrate the rationale of PSHGCN in the graph optimization framework. We conducted an extensive experimental study to show that PSHGCN can learn diverse heterogeneous graph filters and outperform all baselines on open benchmarks. Notably, PSHGCN exhibits remarkable scalability, efficiently handling large real-world graphs comprising millions of nodes and edges. Our codes are available at https://github.com/ivam-he/PSHGCN.
This paper addresses unsupervised representation learning on tabular data containing multiple views generated by distinct sources of measurement. Traditional methods, which tackle this problem using the multi-view framework, are constrained by predefined assumptions that assume feature sets share the same information and representations should learn globally shared factors. However, this assumption is not always valid for real-world tabular datasets with complex dependencies between feature sets, resulting in localized information that is harder to learn. To overcome this limitation, we propose a data-driven approach that learns feature set dependencies by representing feature sets as graph nodes and their relationships as learnable edges. Furthermore, we introduce LEGATO, a novel hierarchical graph autoencoder that learns a smaller, latent graph to aggregate information from multiple views dynamically. This approach results in latent graph components that specialize in capturing localized information from different regions of the input, leading to superior downstream performance.
We introduce Logical Offline Cycle Consistency Optimization (LOCCO), a scalable, semi-supervised method for training a neural semantic parser. Conceptually, LOCCO can be viewed as a form of self-learning where the semantic parser being trained is used to generate annotations for unlabeled text that are then used as new supervision. To increase the quality of annotations, our method utilizes a count-based prior over valid formal meaning representations and a cycle-consistency score produced by a neural text generation model as additional signals. Both the prior and semantic parser are updated in an alternate fashion from full passes over the training data, which can be seen as approximating the marginalization of latent structures through stochastic variational inference. The use of a count-based prior, frozen text generation model, and offline annotation process yields an approach with negligible complexity and latency increases as compared to conventional self-learning. As an added bonus, the annotations produced by LOCCO can be trivially repurposed to train a neural text generation model. We demonstrate the utility of LOCCO on the well-known WebNLG benchmark where we obtain an improvement of 2 points against a self-learning parser under equivalent conditions, an improvement of 1.3 points against the previous state-of-the-art parser, and competitive text generation performance in terms of BLEU score.
Despite the importance of denoising in modern machine learning and ample empirical work on supervised denoising, its theoretical understanding is still relatively scarce. One concern about studying supervised denoising is that one might not always have noiseless training data from the test distribution. It is more reasonable to have access to noiseless training data from a different dataset than the test dataset. Motivated by this, we study supervised denoising and noisy-input regression under distribution shift. We add three considerations to increase the applicability of our theoretical insights to real-life data and modern machine learning. First, while most past theoretical work assumes that the data covariance matrix is full-rank and well-conditioned, empirical studies have shown that real-life data is approximately low-rank. Thus, we assume that our data matrices are low-rank. Second, we drop independence assumptions on our data. Third, the rise in computational power and dimensionality of data have made it important to study non-classical regimes of learning. Thus, we work in the non-classical proportional regime, where data dimension $d$ and number of samples $N$ grow as $d/N = c + o(1)$. For this setting, we derive data-dependent, instance specific expressions for the test error for both denoising and noisy-input regression, and study when overfitting the noise is benign, tempered or catastrophic. We show that the test error exhibits double descent under general distribution shift, providing insights for data augmentation and the role of noise as an implicit regularizer. We also perform experiments using real-life data, where we match the theoretical predictions with under 1\% MSE error for low-rank data.
Much of the recent work developing formal methods techniques to specify or learn the behavior of autonomous systems is predicated on a belief that formal specifications are interpretable and useful for humans when checking systems. Though frequently asserted, this assumption is rarely tested. We performed a human experiment (N = 62) with a mix of people who were and were not familiar with formal methods beforehand, asking them to validate whether a set of signal temporal logic (STL) constraints would keep an agent out of harm and allow it to complete a task in a gridworld capture-the-flag setting. Validation accuracy was $45\% \pm 20\%$ (mean $\pm$ standard deviation). The ground-truth validity of a specification, subjects' familiarity with formal methods, and subjects' level of education were found to be significant factors in determining validation correctness. Participants exhibited an affirmation bias, causing significantly increased accuracy on valid specifications, but significantly decreased accuracy on invalid specifications. Additionally, participants, particularly those familiar with formal methods, tended to be overconfident in their answers, and be similarly confident regardless of actual correctness. Our data do not support the belief that formal specifications are inherently human-interpretable to a meaningful degree for system validation. We recommend ergonomic improvements to data presentation and validation training, which should be tested before claims of interpretability make their way back into the formal methods literature.
Recent advancements in conversational large language models (LLMs), such as ChatGPT, have demonstrated remarkable promise in various domains, including drug discovery. However, existing works mainly focus on investigating the capabilities of conversational LLMs on chemical reaction and retrosynthesis. While drug editing, a critical task in the drug discovery pipeline, remains largely unexplored. To bridge this gap, we propose ChatDrug, a framework to facilitate the systematic investigation of drug editing using LLMs. ChatDrug jointly leverages a prompt module, a retrieval and domain feedback (ReDF) module, and a conversation module to streamline effective drug editing. We empirically show that ChatDrug reaches the best performance on 33 out of 39 drug editing tasks, encompassing small molecules, peptides, and proteins. We further demonstrate, through 10 case studies, that ChatDrug can successfully identify the key substructures (e.g., the molecule functional groups, peptide motifs, and protein structures) for manipulation, generating diverse and valid suggestions for drug editing. Promisingly, we also show that ChatDrug can offer insightful explanations from a domain-specific perspective, enhancing interpretability and enabling informed decision-making. This research sheds light on the potential of ChatGPT and conversational LLMs for drug editing. It paves the way for a more efficient and collaborative drug discovery pipeline, contributing to the advancement of pharmaceutical research and development.
Selective rationales and counterfactual examples have emerged as two effective, complementary classes of interpretability methods for analyzing and training NLP models. However, prior work has not explored how these methods can be integrated to combine their complementary advantages. We overcome this limitation by introducing CREST (ContRastive Edits with Sparse raTionalization), a joint framework for selective rationalization and counterfactual text generation, and show that this framework leads to improvements in counterfactual quality, model robustness, and interpretability. First, CREST generates valid counterfactuals that are more natural than those produced by previous methods, and subsequently can be used for data augmentation at scale, reducing the need for human-generated examples. Second, we introduce a new loss function that leverages CREST counterfactuals to regularize selective rationales and show that this regularization improves both model robustness and rationale quality, compared to methods that do not leverage CREST counterfactuals. Our results demonstrate that CREST successfully bridges the gap between selective rationales and counterfactual examples, addressing the limitations of existing methods and providing a more comprehensive view of a model's predictions.