Aaron Steven White

Yunmo Chen, William Gantt, Weiwei Gu et al. · microsoft-research

We present a novel iterative extraction model, IterX, for extracting complex relations, or templates (i.e., N-tuples representing a mapping from named slots to spans of text) within a document. Documents may feature zero or more instances of a template of any given type, and the task of template extraction entails identifying the templates in a document and extracting each template's slot values. Our imitation learning approach casts the problem as a Markov decision process (MDP), and relieves the need to use predefined template orders to train an extractor. It leads to state-of-the-art results on two established benchmarks -- 4-ary relation extraction on SciREX and template extraction on MUC-4 -- as well as a strong baseline on the new BETTER Granular task.

Yunmo Chen, William Gantt, Tongfei Chen et al. · microsoft-research

We present a conceptual framework that unifies a variety of evaluation metrics for different structured prediction tasks (e.g. event and relation extraction, syntactic and semantic parsing). Our framework requires representing the outputs of these tasks as objects of certain data types, and derives metrics through matching of common substructures, possibly followed by normalization. We demonstrate how commonly used metrics for a number of tasks can be succinctly expressed by this framework, and show that new metrics can be naturally derived in a bottom-up way based on an output structure. We release a library that enables this derivation to create new metrics. Finally, we consider how specific characteristics of tasks motivate metric design decisions, and suggest possible modifications to existing metrics in line with those motivations.

Samuel Barham, Orion Weller, Michelle Yuan et al.

To foster the development of new models for collaborative AI-assisted report generation, we introduce MegaWika, consisting of 13 million Wikipedia articles in 50 diverse languages, along with their 71 million referenced source materials. We process this dataset for a myriad of applications, going beyond the initial Wikipedia citation extraction and web scraping of content, including translating non-English articles for cross-lingual applications and providing FrameNet parses for automated semantic analysis. MegaWika is the largest resource for sentence-level report generation and the only report generation dataset that is multilingual. We manually analyze the quality of this resource through a semantically stratified sample. Finally, we provide baseline results and trained models for crucial steps in automated report generation: cross-lingual question answering and citation retrieval.

Siddharth Vashishtha, Alexander Martin, William Gantt et al.

Understanding event descriptions is a central aspect of language processing, but current approaches focus overwhelmingly on single sentences or documents. Aggregating information about an event \emph{across documents} can offer a much richer understanding. To this end, we present FAMuS, a new corpus of Wikipedia passages that \emph{report} on some event, paired with underlying, genre-diverse (non-Wikipedia) \emph{source} articles for the same event. Events and (cross-sentence) arguments in both report and source are annotated against FrameNet, providing broad coverage of different event types. We present results on two key event understanding tasks enabled by FAMuS: \emph{source validation} -- determining whether a document is a valid source for a target report event -- and \emph{cross-document argument extraction} -- full-document argument extraction for a target event from both its report and the correct source article. We release both FAMuS and our models to support further research.

William Gantt, Reno Kriz, Yunmo Chen et al.

As information extraction (IE) systems have grown more adept at processing whole documents, the classic task of template filling has seen renewed interest as benchmark for document-level IE. In this position paper, we call into question the suitability of template filling for this purpose. We argue that the task demands definitive answers to thorny questions of event individuation -- the problem of distinguishing distinct events -- about which even human experts disagree. Through an annotation study and error analysis, we show that this raises concerns about the usefulness of template filling metrics, the quality of datasets for the task, and the ability of models to learn it. Finally, we consider possible solutions.

William Gantt, Shabnam Behzad, Hannah YoungEun An et al.

We introduce MultiMUC, the first multilingual parallel corpus for template filling, comprising translations of the classic MUC-4 template filling benchmark into five languages: Arabic, Chinese, Farsi, Korean, and Russian. We obtain automatic translations from a strong multilingual machine translation system and manually project the original English annotations into each target language. For all languages, we also provide human translations for sentences in the dev and test splits that contain annotated template arguments. Finally, we present baselines on MultiMUC both with state-of-the-art template filling models and with ChatGPT.

William Gantt, Alexander Martin, Pavlo Kuchmiichuk et al.

We introduce event-keyed summarization (EKS), a novel task that marries traditional summarization and document-level event extraction, with the goal of generating a contextualized summary for a specific event, given a document and an extracted event structure. We introduce a dataset for this task, MUCSUM, consisting of summaries of all events in the classic MUC-4 dataset, along with a set of baselines that comprises both pretrained LM standards in the summarization literature, as well as larger frontier models. We show that ablations that reduce EKS to traditional summarization or structure-to-text yield inferior summaries of target events and that MUCSUM is a robust benchmark for this task. Lastly, we conduct a human evaluation of both reference and model summaries, and provide some detailed analysis of the results.

Angela Cao, Faye Holt, Jonas Chan et al.

With the goal of supporting scalable lexical semantic annotation, analysis, and theorizing, we conduct a comprehensive evaluation of different methods for generating event descriptions under both syntactic constraints -- e.g. desired clause structure -- and semantic constraints -- e.g. desired verb sense. We compare three different methods -- (i) manual generation by experts; (ii) sampling from a corpus annotated for syntactic and semantic information; and (iii) sampling from a language model (LM) conditioned on syntactic and semantic information -- along three dimensions of the generated event descriptions: (a) naturalness, (b) typicality, and (c) distinctiveness. We find that all methods reliably produce natural, typical, and distinctive event descriptions, but that manual generation continues to produce event descriptions that are more natural, typical, and distinctive than the automated generation methods. We conclude that the automated methods we consider produce event descriptions of sufficient quality for use in downstream annotation and analysis insofar as the methods used for this annotation and analysis are robust to a small amount of degradation in the resulting event descriptions.

William Walden, Pavlo Kuchmiichuk, Alexander Martin et al.

Event-keyed summarization (EKS) requires summarizing a specific event described in a document given the document text and an event representation extracted from it. In this work, we extend EKS to the cross-document setting (CDEKS), in which summaries must synthesize information from accounts of the same event as given by multiple sources. We introduce SEAMUS (Summaries of Events Across Multiple Sources), a high-quality dataset for CDEKS based on an expert reannotation of the FAMUS dataset for cross-document argument extraction. We present a suite of baselines on SEAMUS -- covering both smaller, fine-tuned models, as well as zero- and few-shot prompted LLMs -- along with detailed ablations and a human evaluation study, showing SEAMUS to be a valuable benchmark for this new task.

William Gantt, Aaron Steven White

Effective ontology transfer has been a major goal of recent work on event argument extraction (EAE). Two methods in particular -- question answering (QA) and template infilling (TI) -- have emerged as promising approaches to this problem. However, detailed explorations of these techniques' ability to actually enable this transfer are lacking. In this work, we provide such a study, exploring zero-shot transfer using both techniques on six major EAE datasets at both the sentence and document levels. Further, we challenge the growing reliance on LLMs for zero-shot extraction, showing that vastly smaller models trained on an appropriate source ontology can yield zero-shot performance superior to that of GPT-3.5 or GPT-4.

Mahsa Yarmohammadi, Shijie Wu, Marc Marone et al.

Zero-shot cross-lingual information extraction (IE) describes the construction of an IE model for some target language, given existing annotations exclusively in some other language, typically English. While the advance of pretrained multilingual encoders suggests an easy optimism of "train on English, run on any language", we find through a thorough exploration and extension of techniques that a combination of approaches, both new and old, leads to better performance than any one cross-lingual strategy in particular. We explore techniques including data projection and self-training, and how different pretrained encoders impact them. We use English-to-Arabic IE as our initial example, demonstrating strong performance in this setting for event extraction, named entity recognition, part-of-speech tagging, and dependency parsing. We then apply data projection and self-training to three tasks across eight target languages. Because no single set of techniques performs the best across all tasks, we encourage practitioners to explore various configurations of the techniques described in this work when seeking to improve on zero-shot training.

Elias Stengel-Eskin, Kenton Murray, Sheng Zhang et al.

While numerous attempts have been made to jointly parse syntax and semantics, high performance in one domain typically comes at the price of performance in the other. This trade-off contradicts the large body of research focusing on the rich interactions at the syntax-semantics interface. We explore multiple model architectures which allow us to exploit the rich syntactic and semantic annotations contained in the Universal Decompositional Semantics (UDS) dataset, jointly parsing Universal Dependencies and UDS to obtain state-of-the-art results in both formalisms. We analyze the behaviour of a joint model of syntax and semantics, finding patterns supported by linguistic theory at the syntax-semantics interface. We then investigate to what degree joint modeling generalizes to a multilingual setting, where we find similar trends across 8 languages.

William Gantt, Lelia Glass, Aaron Steven White

We present an event structure classification empirically derived from inferential properties annotated on sentence- and document-level Universal Decompositional Semantics (UDS) graphs. We induce this classification jointly with semantic role, entity, and event-event relation classifications using a document-level generative model structured by these graphs. To support this induction, we augment existing annotations found in the UDS1.0 dataset, which covers the entirety of the English Web Treebank, with an array of inferential properties capturing fine-grained aspects of the temporal and aspectual structure of events. The resulting dataset (available at decomp.io) is the largest annotation of event structure and (partial) event coreference to date.

Haoran Xu, Seth Ebner, Mahsa Yarmohammadi et al.

Fine-tuning is known to improve NLP models by adapting an initial model trained on more plentiful but less domain-salient examples to data in a target domain. Such domain adaptation is typically done using one stage of fine-tuning. We demonstrate that gradually fine-tuning in a multi-stage process can yield substantial further gains and can be applied without modifying the model or learning objective.

Patrick Xia, Guanghui Qin, Siddharth Vashishtha et al.

We present LOME, a system for performing multilingual information extraction. Given a text document as input, our core system identifies spans of textual entity and event mentions with a FrameNet (Baker et al., 1998) parser. It subsequently performs coreference resolution, fine-grained entity typing, and temporal relation prediction between events. By doing so, the system constructs an event and entity focused knowledge graph. We can further apply third-party modules for other types of annotation, like relation extraction. Our (multilingual) first-party modules either outperform or are competitive with the (monolingual) state-of-the-art. We achieve this through the use of multilingual encoders like XLM-R (Conneau et al., 2020) and leveraging multilingual training data. LOME is available as a Docker container on Docker Hub. In addition, a lightweight version of the system is accessible as a web demo.

William Gantt, Benjamin Kane, Aaron Steven White

There is growing evidence that the prevalence of disagreement in the raw annotations used to construct natural language inference datasets makes the common practice of aggregating those annotations to a single label problematic. We propose a generic method that allows one to skip the aggregation step and train on the raw annotations directly without subjecting the model to unwanted noise that can arise from annotator response biases. We demonstrate that this method, which generalizes the notion of a \textit{mixed effects model} by incorporating \textit{annotator random effects} into any existing neural model, improves performance over models that do not incorporate such effects.

Gene Louis Kim, Aaron Steven White

We propose a computational modeling framework for inducing combinatory categorial grammars from arbitrary behavioral data. This framework provides the analyst fine-grained control over the assumptions that the induced grammar should conform to: (i) what the primitive types are; (ii) how complex types are constructed; (iii) what set of combinators can be used to combine types; and (iv) whether (and to what) the types of some lexical items should be fixed. In a proof-of-concept experiment, we deploy our framework for use in distributional analysis. We focus on the relationship between s(emantic)-selection and c(ategory)-selection, using as input a lexicon-scale acceptability judgment dataset focused on English verbs' syntactic distribution (the MegaAcceptability dataset) and enforcing standard assumptions from the semantics literature on the induced grammar.

Aaron Steven White, Kyle Rawlins

We investigate the relationship between the frequency with which verbs are found in particular subcategorization frames and the acceptability of those verbs in those frames, focusing in particular on subordinate clause-taking verbs, such as "think", "want", and "tell". We show that verbs' subcategorization frame frequency distributions are poor predictors of their acceptability in those frames---explaining, at best, less than 1/3 of the total information about acceptability across the lexicon---and, further, that common matrix factorization techniques used to model the acquisition of verbs' acceptability in subcategorization frames fare only marginally better. All data and code are available at http://megaattitude.io.

Yunmo Chen, Tongfei Chen, Seth Ebner et al.

We ask whether text understanding has progressed to where we may extract event information through incremental refinement of bleached statements derived from annotation manuals. Such a capability would allow for the trivial construction and extension of an extraction framework by intended end-users through declarations such as, "Some person was born in some location at some time." We introduce an example of a model that employs such statements, with experiments illustrating we can extract events under closed ontologies and generalize to unseen event types simply by reading new definitions.

Elias Stengel-Eskin, Aaron Steven White, Sheng Zhang et al.

We introduce a transductive model for parsing into Universal Decompositional Semantics (UDS) representations, which jointly learns to map natural language utterances into UDS graph structures and annotate the graph with decompositional semantic attribute scores. We also introduce a strong pipeline model for parsing into the UDS graph structure, and show that our transductive parser performs comparably while additionally performing attribute prediction. By analyzing the attribute prediction errors, we find the model captures natural relationships between attribute groups.

Aaron Steven White, Elias Stengel-Eskin, Siddharth Vashishtha et al.

We present the Universal Decompositional Semantics (UDS) dataset (v1.0), which is bundled with the Decomp toolkit (v0.1). UDS1.0 unifies five high-quality, decompositional semantics-aligned annotation sets within a single semantic graph specification---with graph structures defined by the predicative patterns produced by the PredPatt tool and real-valued node and edge attributes constructed using sophisticated normalization procedures. The Decomp toolkit provides a suite of Python 3 tools for querying UDS graphs using SPARQL. Both UDS1.0 and Decomp0.1 are publicly available at http://decomp.io.

Hannah Youngeun An, Aaron Steven White

We investigate neg(ation)-raising inferences, wherein negation on a predicate can be interpreted as though in that predicate's subordinate clause. To do this, we collect a large-scale dataset of neg-raising judgments for effectively all English clause-embedding verbs and develop a model to jointly induce the semantic types of verbs and their subordinate clauses and the relationship of these types to neg-raising inferences. We find that some neg-raising inferences are attributable to properties of particular predicates, while others are attributable to subordinate clause structure.

Shaorong Yan, Aaron Steven White

We propose a novel framework for modeling event-related potentials (ERPs) collected during reading that couples pre-trained convolutional decoders with a language model. Using this framework, we compare the abilities of a variety of existing and novel sentence processing models to reconstruct ERPs. We find that modern contextual word embeddings underperform surprisal-based models but that, combined, the two outperform either on its own.

Siddharth Vashishtha, Benjamin Van Durme, Aaron Steven White

We present a novel semantic framework for modeling temporal relations and event durations that maps pairs of events to real-valued scales. We use this framework to construct the largest temporal relations dataset to date, covering the entirety of the Universal Dependencies English Web Treebank. We use this dataset to train models for jointly predicting fine-grained temporal relations and event durations. We report strong results on our data and show the efficacy of a transfer-learning approach for predicting categorical relations.

Venkata Subrahmanyan Govindarajan, Benjamin Van Durme, Aaron Steven White

We present a novel semantic framework for modeling linguistic expressions of generalization---generic, habitual, and episodic statements---as combinations of simple, real-valued referential properties of predicates and their arguments. We use this framework to construct a dataset covering the entirety of the Universal Dependencies English Web Treebank. We use this dataset to probe the efficacy of type-level and token-level information---including hand-engineered features and static (GloVe) and contextual (ELMo) word embeddings---for predicting expressions of generalization. Data and code are available at decomp.io.

Aaron Steven White, Rachel Rudinger, Kyle Rawlins et al.

We investigate neural models' ability to capture lexicosyntactic inferences: inferences triggered by the interaction of lexical and syntactic information. We take the task of event factuality prediction as a case study and build a factuality judgment dataset for all English clause-embedding verbs in various syntactic contexts. We use this dataset, which we make publicly available, to probe the behavior of current state-of-the-art neural systems, showing that these systems make certain systematic errors that are clearly visible through the lens of factuality prediction.

Adam Poliak, Aparajita Haldar, Rachel Rudinger et al.

We present a large-scale collection of diverse natural language inference (NLI) datasets that help provide insight into how well a sentence representation captures distinct types of reasoning. The collection results from recasting 13 existing datasets from 7 semantic phenomena into a common NLI structure, resulting in over half a million labeled context-hypothesis pairs in total. We refer to our collection as the DNC: Diverse Natural Language Inference Collection. The DNC is available online at https://www.decomp.net, and will grow over time as additional resources are recast and added from novel sources.

Rachel Rudinger, Aaron Steven White, Benjamin Van Durme

We present two neural models for event factuality prediction, which yield significant performance gains over previous models on three event factuality datasets: FactBank, UW, and MEANTIME. We also present a substantial expansion of the It Happened portion of the Universal Decompositional Semantics dataset, yielding the largest event factuality dataset to date. We report model results on this extended factuality dataset as well.

Aaron Steven White, Drew Reisinger, Rachel Rudinger et al.

A linking theory explains how verbs' semantic arguments are mapped to their syntactic arguments---the inverse of the Semantic Role Labeling task from the shallow semantic parsing literature. In this paper, we develop the Computational Linking Theory framework as a method for implementing and testing linking theories proposed in the theoretical literature. We deploy this framework to assess two cross-cutting types of linking theory: local v. global models and categorical v. featural models. To further investigate the behavior of these models, we develop a measurement model in the spirit of previous work in semantic role induction: the Semantic Proto-Role Linking Model. We use this model, which implements a generalization of Dowty's seminal Proto-Role Theory, to induce semantic proto-roles, which we compare to those Dowty proposes.